JP6472327B2 - Abnormal sign diagnostic apparatus and abnormal sign diagnostic method - Google Patents

Description

  The present invention relates to an abnormality sign diagnostic apparatus for diagnosing the presence or absence of an abnormality sign of mechanical equipment.

  2. Description of the Related Art A technique for diagnosing the presence / absence of abnormal signs of mechanical equipment based on detection values of sensors installed in mechanical equipment such as diesel engines, gas engines, and chemical plants is known.

  For example, Patent Literature 1 discloses a learning target data acquisition process for acquiring a detection value of a sensor installed in a mechanical facility as learning target data, a learning process for learning a cluster indicating a normal range of the learning target data, and the cluster. And a diagnostic process for diagnosing the presence / absence of an abnormality sign of mechanical equipment based on the above.

JP2013-008092A

  By the way, depending on the type of mechanical equipment (for example, a diesel engine), even though the mechanical equipment is normal, the state (that is, the detection value of the sensor) fluctuates greatly, or the state fluctuates frequently. Sometimes. As a result, the above-described learning target data includes data having a very large detection value of the sensor installed in the mechanical facility and data having a very small detection value, thereby greatly increasing the radius of the cluster. This may increase the accuracy of diagnosis.

  Then, this invention makes it a subject to provide the abnormality sign diagnostic apparatus etc. which can diagnose appropriately the presence or absence of the abnormality sign of a mechanical installation.

In order to solve the above problems, an abnormality sign diagnosis apparatus according to the present invention includes a detection value of a first sensor installed in a mechanical facility, detection values of a plurality of types of second sensors installed in the mechanical facility, Based on time series data acquisition means for acquiring time series data including: past time series data in which the mechanical equipment is known to be normal; and a detection value of the first sensor and a detection of the first sensor Storage means for storing a function representing an approximate curve of a distribution of points specified by the detection value of the second sensor correlated with the value, and a plurality of types of the second sensors; Correction means for correcting time-series data based on the function stored in the means, and diagnosis means for diagnosing the presence or absence of an abnormality sign of the mechanical equipment, wherein the correction means is detected by the second sensor. A predetermined reference value from the value Calculating a second difference by calculating, and calculating a first difference corresponding to the second difference with respect to the first sensor based on the function for each of the plurality of types of the second sensors, A correction value related to the first sensor is calculated by subtracting a first difference integrated value, which is the sum of the first differences, from the detection value of the first sensor, and the diagnostic means determines that the correction value is out of a predetermined range. In this case, it is diagnosed that there is a sign of abnormality in the mechanical equipment, and the correction means performs the first sensor based on the detection values and the functions of a plurality of types of the second sensor as preprocessing of diagnosis by the diagnosis means. A process of calculating the correction value relating to is performed .

  ADVANTAGE OF THE INVENTION According to this invention, the abnormal sign diagnostic apparatus etc. which diagnose appropriately the presence or absence of the abnormal sign of mechanical equipment can be provided.

It is a block diagram of the abnormality sign diagnostic apparatus which concerns on 1st Embodiment of this invention. It is explanatory drawing of an internal combustion power generator. (A) is a scatter diagram in which points specified by the intake air temperature and exhaust temperature of the internal combustion power generator are plotted, and (b) is specified by the power generation output of the internal combustion power generator and the exhaust temperature. It is a scatter diagram in which points are plotted. It is a flowchart regarding derivation | leading-out of the function showing an approximated curve. It is a flowchart regarding correction | amendment of time series data. (A) is explanatory drawing regarding correction | amendment of exhaust temperature, (b) is explanatory drawing which shows the relationship between the 2nd difference regarding intake temperature, and the 1st difference regarding exhaust temperature, (c) is related with generated electric power. It is explanatory drawing which shows the relationship between a 2nd difference and the 1st difference regarding exhaust gas temperature. It is a flowchart regarding the diagnosis of the presence or absence of an abnormality sign. (A) is the experimental data regarding exhaust temperature, (b) is the elements on larger scale of the area | region K shown to (a). (A) is experimental data regarding the correction value of exhaust temperature, (b) is the elements on larger scale of the area | region M shown to (a). It is a block diagram of the abnormality sign diagnostic apparatus which concerns on 2nd Embodiment. It is a block diagram of the data mining means with which an abnormality sign diagnostic apparatus is provided. It is explanatory drawing of the cluster learned by a cluster learning part. It is a flowchart of the learning process which a learning means performs. It is a flowchart of the diagnostic process which a diagnostic means performs. It is explanatory drawing which shows the several cluster learned by the learning means, and the feature vector of each time series data.

<< First Embodiment >>
FIG. 1 is a configuration diagram of an abnormality sign diagnosis apparatus 1 according to the first embodiment.
The abnormality sign diagnosis device 1 detects an abnormality sign in the internal combustion power generator 2 based on time-series data including detection values of a plurality of sensors 31 to 33 (see FIG. 2) installed in the internal combustion power generator 2 (mechanical equipment). This is a device for diagnosing whether or not an error has occurred. The above-mentioned “abnormal sign” is a prelude to the occurrence of an abnormality in the internal combustion power generator 2, and “abnormal sign diagnosis” is to diagnose the presence or absence of an abnormal sign.
In the following, prior to the description of the abnormality sign diagnosis apparatus 1, the internal combustion power generator 2 that is the object of abnormality sign diagnosis and the sensors 31 to 33 (see FIG. 2) installed in the internal combustion power generator 2 will be described. Briefly described.

<Configuration of internal combustion power generator>
FIG. 2 is an explanatory diagram of the internal combustion power generator 2. The internal combustion power generator 2 is a device that converts chemical energy into kinetic energy by the internal combustion engine 21 and converts the kinetic energy into electrical energy by the generator 27. Below, the case where the internal combustion engine 21 is a diesel engine is demonstrated as an example.

The internal combustion power generator 2 includes an internal combustion engine 21, an air supply passage 22, an air supply pump 23, a fuel gas supply passage 24, a fuel gas supply pump 25, a gas discharge passage 26, and a generator 27. And.
Although not shown, the internal combustion engine 21 compresses air and fuel gas supplied into the cylinder, causes the piston to reciprocate in the cylinder by expansion accompanying spontaneous ignition of the fuel gas, and rotates the crankshaft by reciprocation of the piston. This is a diesel engine configured as described above.

The air supply flow path 22 is a flow path that guides air to a combustion chamber (not shown) of the internal combustion engine 21. The air supply pump 23 pumps air toward the combustion chamber, and is installed in the air supply flow path 22.
The fuel gas supply channel 24 is a channel that guides the fuel gas to the combustion chamber of the internal combustion engine 21. The fuel gas supply pump 25 pumps fuel gas toward the combustion chamber, and is installed in the fuel gas supply flow path 24.

The gas discharge channel 26 is a channel for discharging the gas burned in the internal combustion engine 21.
The generator 27 generates power by driving the internal combustion engine 21, and a rotor (not shown) is connected to a crankshaft (not shown) of the internal combustion engine 21.

<Sensor>
The intake air temperature sensor 31 (second sensor) is a sensor that detects the temperature (intake air temperature) of air supplied to the internal combustion engine 21, and is installed in the air supply flow path 22.
The exhaust gas temperature sensor 32 (first sensor) is a sensor that detects the temperature (exhaust gas temperature) of the gas discharged from the internal combustion engine 21, and is installed in the gas discharge channel 26.
The generated power sensor 33 (second sensor) is a sensor that detects the generated power of the generator 27.

  Here, the “first sensor” is a sensor whose detection value is directly used for diagnosis of an abnormal sign of the internal combustion power generator 2. In the present embodiment, the exhaust temperature sensor 32 corresponds to a “first sensor”. For example, when the cooling water supplied to the internal combustion engine 21 is insufficient or the lubricating oil is insufficient, the temperature of the internal combustion engine 21 becomes abnormally high, and the temperature of the gas discharged from the combustion chamber (not shown) is also high. Get higher. That is, the detected value of the exhaust gas temperature sensor 32 sensitively reflects the state of the internal combustion power generator 2 (presence / absence of an abnormal sign), so in this embodiment, an abnormal sign diagnosis is performed based on the exhaust temperature. Yes.

  The “second sensor” is a sensor having a correlation between the detected value and the detected value of the “first sensor” when the internal combustion power generator 2 is normal. In the present embodiment, the intake air temperature sensor 31 and the generated power sensor 33 correspond to a “second sensor”.

For example, although the internal combustion power generator 2 is normal, the detected value of the exhaust gas temperature sensor 32 used for diagnosing abnormal signs is also accompanied by fluctuations in the command value of the generated power and fluctuations in the outside air temperature (intake air temperature). Often fluctuates greatly.
Conventionally, an abnormality sign diagnosis is performed by comparing the detected value of the exhaust temperature sensor 32 with a predetermined threshold value. However, as described above, the detection of the exhaust temperature sensor 32 is performed even when the internal combustion power generator 2 is normal. Since the value fluctuated greatly, there was a possibility of causing a misdiagnosis.

On the other hand, in the present embodiment, a function representing the correlation between the detected value of the exhaust temperature sensor 32 and the detected value of the intake air temperature sensor 31, or the detected value of the exhaust temperature sensor 32 and the detected value of the generated power sensor 33 The detection value of the exhaust temperature sensor 32 is corrected using a function representing the correlation. These functions will be described later.
The detection values of the sensors 31 to 33 shown in FIG. 2 are transmitted to the abnormality sign diagnosis apparatus 1 at a predetermined cycle via the communication unit 34 and the network N.

  In FIG. 1, the configuration in which the detection values of the sensors 31 to 33 are transmitted to the abnormality sign diagnosis apparatus 1 via the network N for each of the plurality of internal combustion power generators 2 is illustrated. The number of may be one. Further, it is not necessary that the plurality of internal combustion power generators 2 have the same configuration.

<Configuration of abnormal sign diagnosis device>
As shown in FIG. 1, the abnormality sign diagnosis apparatus 1 includes a communication unit 11, a time series data acquisition unit 12, a time series data storage unit 13, a function derivation unit 14, a function storage unit 15, and a correction unit 16. And a corrected time-series data storage unit 17, a diagnosis unit 18, a diagnosis result storage unit 19, and a display control unit 20.

The communication means 11 receives data transmitted from the internal combustion power generator 2 via the network N based on a predetermined communication protocol.
The time-series data acquisition unit 12 acquires time-series data related to the internal combustion power generator 2 among the data received by the communication unit 11 via the network N. In the time series data described above, the physical quantity is determined by the identification information of the internal combustion power generator 2, the detection values of the sensors 31 to 33 (see FIG. 2), the identification information of the sensors 31 to 33, and the sensors 31 to 33. The detected date and time are included.

The time series data storage unit 13 stores the time series data acquired by the time series data acquisition unit 12 as, for example, a database.
A magnetic disk device, an optical disk device, a semiconductor storage device, or the like can be used as the time series data storage means 13 (the function storage means 15, the corrected time series data storage means 17, and the diagnosis result storage means 19 described later). The same).

The function deriving means 14 is based on the past time series data that the internal combustion power generator 2 is known to be normal, and the detected value of the intake air temperature sensor 31 (see FIG. 2) and the exhaust gas temperature sensor 32 (see FIG. 2). ) And a function for deriving a function representing an approximate curve of the distribution of points specified by
Further, the function deriving unit 14 detects the detected value of the exhaust gas temperature sensor 32 (see FIG. 2) and the generated power sensor 33 (see FIG. 2) based on past time series data in which the internal combustion power generator 2 is known to be normal. 2) and a function for deriving a function representing an approximate curve of the distribution of points specified by
The “approximate curve” includes an “approximate straight line”.

FIG. 3A is a scatter diagram in which points specified by the intake air temperature and the exhaust gas temperature of the internal combustion power generator 2 are plotted. The horizontal axis of this scatter diagram is the intake air temperature detected by the intake air temperature sensor 31 (see FIG. 2), and the vertical axis is the exhaust gas temperature detected by the exhaust gas temperature sensor 32 (see FIG. 2).
In addition, each point shown to Fig.3 (a) is when it is known that the internal combustion power generator 2 is normal, and the internal combustion power generator 2 is operated so that predetermined generation power can be obtained. The average value for every month (for 12 months) is calculated for each of the detection values obtained.

  As shown in FIG. 3A, when the internal combustion power generator 2 is normal, there is a strong correlation between the intake air temperature and the exhaust gas temperature, and the exhaust gas temperature increases as the intake air temperature increases. ing. The presence or absence of such correlation is specified in advance based on the results of prior experiments and simulations.

The function deriving means 14 shown in FIG. 2 derives a function representing an approximate curve of the distribution of each point specified by the detected value when the intake air temperature sensor 31 and the exhaust gas temperature sensor 32 are designated by the operation of the computer 4. . As a method for deriving such an approximate curve, for example, a least square method can be used. In the example shown in FIG. 3A, a function y = ax + b representing the straight line G _A (approximate curve) is derived by the function deriving means 14.

FIG. 3B is a scatter diagram in which points specified by the power generation output of the internal combustion power generator 2 and the exhaust gas temperature are plotted. The horizontal axis of this scatter diagram is the generated power detected by the generated power sensor 33 (see FIG. 2), and the vertical axis is the exhaust temperature detected by the exhaust temperature sensor 32 (see FIG. 2).
Each point shown in FIG. 3B is obtained during a period in which the internal combustion power generator 2 is known to be normal and the intake air temperature (outside air temperature) is substantially constant.

  As shown in FIG. 3 (b), when the internal combustion power generator 2 is normal, there is also a strong correlation between the power generation output and the exhaust temperature, and the larger the generated power, the higher the exhaust temperature. Yes. The presence or absence of such correlation is specified in advance based on the results of prior experiments and simulations.

As described above, when the exhaust temperature sensor 32 and the generated power sensor 33 are designated by the operation of the computer 4, the function deriving unit 14 shown in FIG. 2 calculates an approximate curve of the distribution of each point specified by the detected value. Derive the function to represent. In the example shown in FIG. 3B, the function y = cx + d representing the straight line G _B (approximate curve) is derived by the function deriving unit 14.
Incidentally, since it is only necessary to be able to derive functions representing the straight lines G _A and G _B , it is not necessary to make each point shown in FIG. 3A correspond to each point shown in FIG. .

A function storage unit 15 shown in FIG. 1 (storage means), a function representing a straight line G _A is stored in association with the intake air temperature sensor 31. Further, the function storage unit 15, a function representing a straight line G _B is stored in association with the generated power sensor 33.
The correction unit 16 has a function of correcting the detected value (time series data) of the exhaust temperature based on two functions (y = ax + b, y = cx + d) stored in the function storage unit 15. The processing executed by the correction unit 16 will be described later.
The corrected time-series data storage means 17 stores a correction value of the exhaust temperature used for diagnosis of an abnormal sign.

The diagnosis unit 18 has a function of diagnosing the presence / absence of an abnormality sign of the internal combustion power generator 2 based on the correction value of the exhaust temperature (corrected time series data) stored in the corrected time series data storage unit 17. Have. The processing executed by the diagnosis unit 18 will be described later.
The diagnosis result storage means 19 stores information related to the diagnosis result of the diagnosis means 18. The above-described diagnosis result includes identification information of the internal combustion power generator 2 and presence / absence of an abnormality sign.

  The display control means 20 outputs a control signal for displaying the diagnosis result by the diagnosis means 18 to the display device 5. For example, the display control means 20 displays the diagnosis result (presence / absence of abnormality) on the display device 5 in a matrix format with the identification information of the internal combustion power generator 2 as a row and the date of diagnosis as a column.

FIG. 4 is a flowchart relating to the derivation of the function representing the approximate curve.
In step S101, the abnormality sign diagnosis device 1 acquires time-series data when the internal combustion power generator 2 is normal by the time-series data acquisition unit 12 (time-series data acquisition step). As described above, it is assumed that the internal combustion power generator 2 is operating normally.
In step S102, the abnormality sign diagnosis apparatus 1 stores the time-series data acquired in step S101 in the time-series data storage unit 13.

In step S <b> 103, the abnormality sign diagnosis apparatus 1 derives a function representing the approximate curve by the function deriving unit 14. That is, the abnormality sign diagnosis apparatus 1 functions as a straight line G _A (see FIG. 3A) that approximates the distribution of each point specified by the intake air temperature and the exhaust gas temperature in accordance with the operation of the computer 4 by the user. Is derived. Further, the abnormality sign diagnosis apparatus 1 is a function representing a straight line G _B (see FIG. 3B) that approximates the distribution of each point specified by the power generation output and the exhaust temperature in accordance with the operation of the computer 4 by the user. Is derived.
In step S104, the abnormality sign diagnosis apparatus 1 stores the function derived in step S103 in the function storage unit 15 (storage step).

FIG. 5 is a flowchart regarding correction of time-series data.
In step S <b> 201, the abnormality sign diagnosis apparatus 1 acquires time series data to be diagnosed by the time series data acquisition unit 12. The time series data to be diagnosed includes the intake air temperature detected by the intake air temperature sensor 31 (see FIG. 2), the exhaust gas temperature detected by the exhaust gas temperature sensor 32 (see FIG. 2), and the generated power sensor 33 (see FIG. 2). And the generated power detected by the reference). Based on each of these detection values, the detection value of the exhaust temperature sensor 32 is corrected as described below.

FIG. 6A is an explanatory diagram regarding correction of the exhaust gas temperature.
The horizontal axis of the explanatory diagram shown in FIG. 6A is the time, and the vertical axis is the exhaust temperature of the internal combustion power generator 2. It is assumed that a series of correction processes (S201 to S209) described below have already been performed on the exhaust temperatures detected at times t1, t2, and t3. Here, the detected value T _S of the exhaust temperature at time t4, the detected value T _Q of the intake air temperature at time t4 (see FIG. 6B), and the detected value W _{Q of} generated power (see FIG. 6C). ) And the case of correction based on the above.

  In step S202 of FIG. 5, the abnormality sign diagnosis apparatus 1 selects one from a plurality of “second sensors”. As described above, the “second sensor” includes the intake air temperature sensor 31 (see FIG. 2) and the generated power sensor 33 (see FIG. 2). The abnormality sign diagnosis apparatus 1 selects, for example, the intake air temperature sensor 31 as the “second sensor”.

  In step S203, the abnormality sign diagnosis apparatus 1 calculates a “second difference” by subtracting a predetermined reference value from the detection value of the intake air temperature sensor 31 for the time-series data acquired in step S201.

FIG. 6B is an explanatory diagram showing the relationship between the second difference ΔT2 related to the intake air temperature and the first difference ΔT1 _A related to the exhaust gas temperature. Function representing a straight line G _A shown in FIG. 6 (b) is based on the correlation between the intake air temperature and exhaust temperature, is derived in step S103 (see FIG. 4), previously stored in the function storage means 15 Yes.

In step S203, the abnormality sign diagnosis apparatus 1 calculates a second difference ΔT2 (= T _Q −T _P ) by subtracting a predetermined reference value T _P (for example, 20 ° C.) from the detected value T _Q of the intake air temperature. To do. Reference value T _P described above is, for example, an average value of the intake air temperature throughout the year (outside air temperature) is preset. In the example shown in FIG. 6 (a), the detected value T _Q is higher than the reference value T _P, the second difference ΔT2 has become a positive value.

In step S204, the abnormality sign diagnosis apparatus 1 calculates a first difference ΔT1 _A corresponding to the second difference ΔT2. That is, the abnormality indication diagnostic apparatus 1 reads out a function representing a straight line _{G A} from the function storage unit 15, the gradient a of the straight line _{G A} by multiplying the second difference Delta] T2, the first difference ΔT1 _A (= a × ΔT2 ) Is calculated.
In step S205, the abnormality sign diagnosis apparatus 1 stores the first difference ΔT1 _A calculated in step S204.

In step S206, the abnormality sign diagnosis apparatus 1 determines whether or not another “second sensor” exists. If there is another “second sensor” (S206: Yes), the process of the abnormality sign diagnosis apparatus 1 proceeds to step S202. In the present embodiment, the generated power sensor 33 is present as the “second sensor” in addition to the intake air temperature sensor 31 (S206: Yes).
In the second step S202, the abnormality sign diagnosis apparatus 1 selects the generated power sensor 33 from a plurality of “second sensors”, and further executes the processes of steps S203 to S205.

FIG. 6 (c), the second difference ΔW2 relates generated power is an explanatory diagram showing a first difference Delta] T1 _B relates to an exhaust temperature, a relationship. Function representing a straight line G _B shown in FIG. 6 (c) is based on the correlation between the generated power and the exhaust temperature, is derived in step S103 (see FIG. 4), previously stored in the function storage means 15 Yes.

In the second step S203, the abnormality sign diagnosis apparatus 1 subtracts a predetermined reference value W _P (for example, 500 kW) from the detected value W _Q of the generated power, thereby obtaining the second difference ΔW2 (= W _Q −W _P ). Is calculated. Reference value W _P described above is, for example, a mean value of the generated power when the internal combustion power generator 2 are load operation, are set in advance. In the example shown in FIG. 6 (c), the detection value W _Q is larger than the reference value W _P, the second difference ΔW2 has become a positive value.

In the second step S204, the abnormality sign diagnosis apparatus 1 calculates a first difference ΔT1 _B corresponding to the second difference ΔW2. That is, the abnormality indication diagnostic apparatus 1 reads out a function representing a straight line _{G B} from the function storage means 15, the slope c of the straight line _{G B} By multiplying the second difference Derutadaburyu2, first difference ΔT1 _B (= c × ΔW2 ) Is calculated.
In the second step S205, the abnormality sign diagnosis apparatus 1 stores the first difference ΔT1 _B calculated in step S204.

With the above processing, the first difference ΔT1 _A and ΔT1 _B are calculated for each of the plurality of second sensors (the intake air temperature sensor 31 and the generated power sensor 33) (S206: Yes). Proceed to step S207.

In step S207, the abnormality sign diagnosis apparatus 1 calculates a first difference integrated value (ΔT1 _A + ΔT1 _B ) that is the sum of the first differences ΔT1 _A and ΔT1 _B stored in step S206.
In step S208, the abnormality sign diagnosis apparatus 1 subtracts the first difference integrated value (ΔT1 _A + ΔT1 _B ) from the detected value T _S of the exhaust temperature, thereby correcting the exhaust temperature correction value T _RS (= T _S −ΔT1 _A −. ΔT1 _B ) is calculated.

In other words, the abnormality indication diagnostic apparatus 1, the intake air temperature is the predetermined reference value T _P, and the exhaust when the generated power is assumed to have running internal combustion power generator 2 to a predetermined reference value W _P The temperature (correction value T _RS ) is estimated based on a function of the straight lines G _A and G _B.
As described above, there is a strong correlation between the intake air temperature and the exhaust gas temperature, and there is also a strong correlation between the generated power and the exhaust gas temperature (see FIG. 3). By utilizing these correlations, the exhaust gas temperature is corrected on the assumption that the intake air temperature and the generated power that change from moment to moment are constant. As a result, when the internal combustion power generator 2 is normal, the variation in the correction value _TRS of the exhaust temperature becomes very small (see FIG. 6A).

By the way, regarding the exhaust temperature detected in the winter (T _Q <T _P ) where the outside air temperature is relatively low, and the exhaust temperature detected when the generated power (W _Q <W _P ) smaller than normal is supplied. The exhaust gas temperature detection value is corrected so as to increase the value.

Abnormal sign diagnostic apparatus 1 in step S209 stores the correction value T _RS of the exhaust temperature calculated in step S208 to the corrected time series data storage unit 17, and ends the process relating to correction of the exhaust temperature (END).
Note that the processing (correction step) of steps S202 to S209 described above is executed by the correction means 16 (see FIG. 1) for each time-series data to be diagnosed acquired in step S201.

FIG. 7 is a flowchart regarding diagnosis of presence / absence of an abnormality sign. In addition, the process (diagnosis step) of step S301-S306 demonstrated below is performed by the diagnostic means 18 (refer FIG. 2) with which the abnormality sign diagnostic apparatus 1 is provided. Also, shown in FIG. 7, "START" Sometimes, a straight line _G A, the correction value of the exhaust temperature based on a function representing the _{G B T RS} (see FIG. 6 (a)) is calculated, the correction value _{T RS} is corrected It is assumed that it is stored in the time series data storage means 17.

In step S301, the abnormality sign diagnosis apparatus 1 reads the corrected exhaust gas temperature value T _RS (that is, time-series data to be diagnosed) from the corrected time-series data storage unit 17.
Abnormal sign diagnostic apparatus in step S302 1, the correction value T _RS of the exhaust gas temperature acquired in step S301 it is determined whether out of a predetermined range. The above-mentioned “predetermined range” is a range that serves as a criterion for determining whether or not an abnormality sign has occurred in the internal combustion power generator 2, and is set in advance based on a prior experiment or the like. For example, the abnormality indication diagnostic apparatus 1 determines the correction value T _RS of the exhaust temperature threshold value T _L above is shown in FIG. 6 (a), and, whether out of a predetermined range is equal to or less than the threshold T _H.

If the exhaust gas correction value _TRS is out of the predetermined range in step S302 (S302: Yes), the process of the abnormality sign diagnosis apparatus 1 proceeds to step S303. In step S303, the abnormality sign diagnosis device 1 diagnoses the internal combustion power generator 2 as “abnormal sign”.
On the other hand, if the exhaust gas correction value _TRS is within the predetermined range in step S302 (S302: No), the process of the abnormality sign diagnosis apparatus 1 proceeds to step S304.

In step S304, the abnormality sign diagnosis apparatus 1 determines whether there is another exhaust gas temperature correction value T _RS (that is, time series data to be diagnosed). When the correction value _{T RS} of the exhaust gas temperature is present in the other (S304: Yes), the processing of the abnormality sign diagnostic apparatus 1 proceeds to step S301. On the other hand, when the correction value _{T RS} of the exhaust gas temperature is not in the other (S304: No), the processing of the abnormality sign diagnostic apparatus 1 proceeds to step S305. In step S <b> 305, the abnormality sign diagnosis device 1 diagnoses “no abnormality sign” for the internal combustion power generator 2.

  After performing the process of step S303 or step S305, in step S306, the abnormality sign diagnosis apparatus 1 displays the diagnosis result on the display device 5 (see FIG. 1) by the display control means 20 (see FIG. 1), and processes related to the diagnosis. End (END).

<Effect>
In the present embodiment, with respect to the detected value T _S of the exhaust temperature (see FIG. 6A), the intake air temperature is a predetermined reference value T _P (see FIG. 6B), and the generated power is a predetermined reference. It calculates a correction value T _RS assuming that the value W _{P (see} FIG. 6 (c)), so as to diagnose the presence or absence of the abnormal sign internal combustion power generator 2 on the basis of the correction value T _RS Yes. As described above, if the internal combustion power generator 2 is normal, the variation in the correction value _TRS of the exhaust temperature is very small even when the intake air temperature and the generated power fluctuate greatly (see FIG. 6A). Therefore, it is possible to prevent erroneous diagnosis that “there is a sign of abnormality” even though the internal combustion power generator 2 is normal.
Further, since the detection values of the exhaust temperature sensor 32 (“first sensor”) are corrected based on the detection values of the intake air temperature sensor 31 and the generated power sensor 33 (two “second sensors”), one “second” The presence / absence of an abnormality sign of the internal combustion power generator 2 can be diagnosed with higher accuracy than the correction based on the detection value of the “sensor”.

  FIG. 8A shows experimental data relating to the exhaust gas temperature. Each data shown in FIG. 8A is the exhaust temperature detected at each time in the period from August 1 to September 20. On August 24, the internal combustion power generator 2 was diagnosed with “abnormal sign”, and thereafter maintenance was performed, and operation resumed from August 28.

FIG. 8B is a partially enlarged view of the region K shown in FIG. As shown in FIG. 8 (b), it can be seen that the exhaust gas temperature fluctuates greatly even during the period in which the internal combustion power generator 2 is operating normally (before August 23).
If the exhaust gas temperature before correction is used as it is and the presence / absence of an abnormality sign is diagnosed based on a comparison with a predetermined threshold value, it is erroneously diagnosed as “abnormal sign is present” even though the internal combustion power generator 2 is normal. There is a high possibility. This is because, when the internal combustion power generator 2 is normal, for example, when the actual generated power fluctuates greatly, the exhaust temperature having a strong correlation with the generated power also fluctuates greatly.

  FIG. 9A shows experimental data regarding the correction value of the exhaust gas temperature. Each data shown in FIG. 9A is obtained by correcting each data shown in FIG. 8A (S201 to S209: see FIG. 5). Before the maintenance carried out around August 26, exhaust temperature correction values were concentrated and distributed around 630 ° C., and the variation in data was very small compared to the exhaust temperature shown in FIG. Yes.

FIG. 9B is a partially enlarged view of the region M shown in FIG. Note that the thresholds T _H and T _L shown in FIG. 9B correspond to the thresholds T _H and T _L shown in FIG.
In the example shown in FIG. 9 (b), higher than the correction value is a threshold value T _H of the exhaust gas temperature around August 24, has been diagnosed with "Yes abnormal sign" by the diagnostic means 18. For example, when the cooling water for cooling the internal combustion engine 21 (see FIG. 2) is insufficient, the correlation shown in FIGS. 3 (a) and 3 (b) is lost, and the correction value of the exhaust temperature is the threshold value. higher than T _H, is diagnosed that "there is an abnormal sign". As described above, according to the present embodiment, it is possible to appropriately and accurately diagnose the presence or absence of an abnormality sign related to the internal combustion power generator 2 based on the correction value of the exhaust gas temperature.

<< Second Embodiment >>
The abnormality sign diagnosis apparatus 1A (see FIG. 10) according to the second embodiment includes a data mining means 18A (see FIG. 10) instead of the diagnosis means 18 (see FIG. 1) described in the first embodiment. . Moreover, the point which diagnoses the presence or absence of the abnormal sign of the mechanical installation 6 provided with many sensors with the abnormal sign diagnostic apparatus 1A differs from 1st Embodiment.
Note that a series of processing (S201 to S209: refer to FIG. 5) regarding correction of time-series data is the same as that of the first embodiment. Therefore, a different part from 1st Embodiment is demonstrated and description is abbreviate | omitted about the overlapping part.

FIG. 10 is a configuration diagram of the abnormality sign diagnosis apparatus 1A according to the second embodiment.
The abnormality sign diagnosis apparatus 1A is an apparatus that diagnoses whether or not an abnormality sign has occurred in the mechanical facility 6. Although not shown in the drawings, the machine facility 6 that is a diagnosis target is provided with a large number (for example, several hundreds) of sensors. As described in the first embodiment, these sensors include a “first sensor” that is directly used for diagnosis of a sign of abnormality, and a plurality of detection values that have a correlation with a detection value of the “first sensor”. “Second sensor”.

In the first embodiment, the case where there is one “first sensor” (exhaust temperature sensor 32: see FIG. 2) has been described, but there may be a plurality of “first sensors”. In this case, the function storage means 15 includes a function representing an approximate curve of the distribution of points specified by the detection value of the “first sensor” and the detection value of the “second sensor” as the “first sensor”. And “second sensor”.
There are also sensors that do not belong to either the “first sensor” or the “second sensor” (sensors that have little correlation with the detection values of other sensors). Used for diagnosis.

<Configuration of abnormal sign diagnosis device>
As shown in FIG. 10, the abnormality sign diagnosis apparatus 1A includes a communication unit 11, a time series data acquisition unit 12, a time series data storage unit 13, a function derivation unit 14, a function storage unit 15, and a correction unit 16. And a corrected time series data storage means 17, a data mining means 18A, a diagnosis result storage means 19, and a display control means 20.

The corrected time-series data storage means 17 includes a correction value relating to the “first sensor” and a detection value of a sensor not included in any of the “first sensor” and the “second sensor”. 6 identification information and the detection time of each sensor are stored in association with each other.
Incidentally, it is not necessary to include the detection value of the “second sensor” in the corrected time series data stored in the corrected time series data storage means 17. As described in the first embodiment, the correction value related to the “first sensor” is calculated on the assumption that the detection value of the “second sensor” is a predetermined reference value (a constant value). is there.

  FIG. 11 is a configuration diagram of the data mining means 18A provided in the abnormality sign diagnosis apparatus 1A. As shown in FIG. 11, the data mining unit 18A includes a learning unit 181 and a diagnostic unit 182.

  The learning unit 181 has a function of learning a cluster (normal model) indicating the range of the normal state of the mechanical equipment 6 by performing clustering which is one of statistical data classification methods. The above-described cluster is an area specified by the cluster center c (see FIG. 12) and the cluster radius r (see FIG. 12) in the multidimensional vector space, and in a period when the mechanical equipment 6 is known to be normal. Learning is performed using the acquired time-series data.

As shown in FIG. 11, the learning means 181 includes a learning target data acquisition unit 181a, a cluster learning unit 181b, and a cluster information storage unit 181c.
The learning target data acquisition unit 181 a acquires time series data (learning target data) to be learned from the corrected time series data storage unit 17. This learning target data is data that has been subjected to the correction (S201 to S209: see FIG. 5) described in the first embodiment with respect to time-series data acquired during a period when it is known that the mechanical equipment 6 is normal. It is.

  The cluster learning unit 181b learns a cluster indicating the range of the normal state of the mechanical equipment 6 based on the learning target data acquired by the learning target data acquisition unit 181a. Here, a cluster learned by the cluster learning unit 181b will be described.

FIG. 12 is an explanatory diagram of clusters learned by the cluster learning unit 181b.
The state of the mechanical equipment 6 is represented as a feature vector having a component obtained by normalizing the detection value of each sensor (correction value for “first sensor”) in a multidimensional vector space. Here, “normalization” is a process for making a non-dimensional quantity by dividing the detected value of the sensor by a representative value (average value, standard deviation, etc.) of the sensor so that they can be compared with each other.

  For ease of explanation, FIG. 12 shows each feature vector in a three-dimensional vector space corresponding to three sensors. Normally, there are a plurality of clusters, but only one cluster is shown in FIG. Each point shown in FIG. 12 corresponds to time-series data acquired at each time.

  Hereinafter, as an example, a case will be described in which clustering is performed using the k-average method which is non-hierarchical clustering. The cluster learning unit 181b classifies each feature vector corresponding to time-series data into several representative groups called clusters for each similar feature vector. That is, the cluster learning unit 181b randomly assigns clusters to each feature vector, and calculates the center of each cluster (cluster center c: see FIG. 12) based on the assigned data. The cluster center c is, for example, the center of gravity of a plurality of feature vectors belonging to the cluster (components of the axes α, β, and γ: see FIG. 12).

  Next, the cluster learning unit 181b obtains the distance between the predetermined feature vector and each cluster center c, and reassigns the feature vector to the cluster having the smallest distance. Such processing is executed for all the feature vectors, and the learning processing for the cluster is also finished when the cluster assignment has not changed. In other cases, the cluster center c is recalculated from the newly allocated cluster, and the above processing is repeated. In this way, the cluster learning unit 181b classifies the feature vector corresponding to each time series data into a plurality of clusters.

Further, the cluster learning unit 181b calculates the coordinate value of the cluster center c (see FIG. 12) and the cluster radius r (see FIG. 12) for each cluster. The cluster radius r is, for example, an average value of the distance between the cluster center c and the feature vector. Note that the method for calculating the cluster radius r is not limited to this. For example, the cluster radius r may be calculated based on the variance value of each component of the feature vector.
The cluster information storage unit 181c illustrated in FIG. 11 stores cluster information (cluster center c, cluster radius r) regarding the clusters learned by the cluster learning unit 181b.

  The diagnosis unit 182 has a function of diagnosing the presence / absence of an abnormality sign of the machine facility 6 using the clusters learned by the learning unit 181. The diagnosis unit 182 includes a diagnosis target data acquisition unit 182a, an abnormality measure calculation unit 182b, and a diagnosis unit 182c.

The diagnosis target data acquisition unit 182 a acquires time series data (diagnosis target data) to be diagnosed from the corrected time series data storage unit 17.
The abnormality measure calculation unit 182b calculates the abnormality measure u of the mechanical equipment 6 using the diagnosis target data acquired by the diagnosis target data acquisition unit 182a. First, the abnormality measure calculation unit 182b normalizes (dimensionless) the diagnosis target data, and converts each time-series data into a feature vector.

  Then, the abnormality measure calculation unit 182b identifies the cluster having the cluster center c (see FIG. 12) that is closest to the feature vector of the diagnosis target data, and calculates the abnormality measure u of the diagnosis target data based on this cluster. In other words, the abnormality measure calculation unit 182b uses, for example, (Formula 1) using the distance d (see FIG. 12) from the cluster center c of the identified cluster to the diagnosis target data and the cluster radius r (see FIG. 12). ) To calculate the abnormal measure u.

  u = d / r (Equation 1)

  The abnormality measure calculation unit 182b outputs the calculated abnormality measure u to the diagnosis unit 182c. Further, the abnormality measure calculation unit 182b stores the diagnosis target data and the abnormality measure u in the diagnosis result storage unit 19 in association with each other.

  The diagnosis unit 182c has a function of diagnosing the presence / absence of an abnormality sign related to the mechanical facility 6 based on the abnormality measure u input from the abnormality measure calculation unit 182b. The processing executed by the diagnosis unit 182c will be described later.

<Operation of abnormal sign diagnosis device>
FIG. 13 is a flowchart of the learning process executed by the learning unit 181.
In step S401, the learning unit 181 acquires learning target data from the corrected time-series data storage unit 17 by the learning target data acquisition unit 181a. This learning target data is time-series data acquired during a period in which it is known that the mechanical equipment 6 is normal and further corrected by the correcting means 16.

In step S402, the learning unit 181 learns a cluster representing the range of the normal state of the machine facility 6 by the cluster learning unit 181b (learning step). That is, the learning unit 181 converts learning target data into feature vectors, and clusters the feature vectors.
In step S403, the learning unit 181 stores the learning result of step S402 in the cluster information storage unit 181c, and ends the learning process (END).

FIG. 14 is a flowchart of the diagnostic process executed by the diagnostic unit 182.
In step S501, the diagnosis unit 182 acquires diagnosis target data from the corrected time series data storage unit 17 by the diagnosis target data acquisition unit 182a. This diagnosis target data is time-series data that has been corrected by the correction means 16.
In step S502, the diagnosis unit 182 calculates the abnormality measure u of the diagnosis target data by using the abnormality measure calculation unit 182b. That is, the diagnosis unit 182 calculates the abnormality measure u based on the cluster having the cluster center c closest to the feature vector of the diagnosis target data.

In step S503, the diagnosis unit 182 performs a diagnosis process on diagnosis target data by the diagnosis unit 182c (diagnosis step). That is, the diagnosis unit 182 diagnoses the presence / absence of an abnormality sign related to the mechanical equipment 6 by comparing the magnitude of the abnormality measure u calculated in step S502 with a predetermined threshold value (= 1).
When the abnormality measure u ≦ 1, the diagnosis target data exists in the cluster area (that is, in the range where the mechanical equipment 6 is normal). When the abnormality measure u ≦ 1 for all diagnosis target data, the diagnosis unit 182 diagnoses “no abnormality sign” with respect to the mechanical equipment 6.
Further, when there is at least one diagnosis target data with an abnormality measure u> 1, the diagnosis target data exists outside the cluster area, and therefore the diagnosis means 182 indicates that there is an “abnormal sign” regarding the mechanical equipment 6. Diagnose.

  In step S504, the diagnosis unit 182 stores the diagnosis result of step S503 in the diagnosis result storage unit 19, and ends the diagnosis process (END).

<Effect>
FIG. 15 is an explanatory diagram showing a plurality of clusters D1, D2, D3 learned by the learning means 181 and feature vectors of each time series data. Each point shown in FIG. 15 represents a feature vector of diagnosis target data.
As described in the first embodiment, the correction process (S201 to S209: see FIG. 5) by the correction unit 16 (see FIG. 10) makes the variation of the correction value related to the “first sensor” very small. Based on the learning target data including such correction values, the clusters D1, D2, and D3 are learned. Therefore, the cluster radii r1 _α , r2 _α , and r3 _α of the clusters D1, D2, and D3 indicated by solid lines can be reduced (r1 _α <r1) as compared with the case where correction is not performed (each cluster indicated by a broken line). etc).

  Conventionally, even if the mechanical equipment 6 is normal, learning target data having a very large detection value of the “first sensor” and learning target data having a very small detection value of the “first sensor” are mixed. There was a thing. In this case, since each feature vector is distributed over a wide range in a multidimensional vector space, the cluster radius tends to be large, and an erroneous diagnosis is caused with respect to an abnormal sign (for example, an abnormal sign is actually generated in the machine facility 2). However, there is a possibility that “no abnormal signs” will be diagnosed).

  On the other hand, according to this embodiment, each feature vector is distributed over a wide range by correcting the detection value of the “first sensor” based on the correlation with the detection value of the “second sensor”. This can be suppressed. Therefore, the normal state of the mechanical equipment 6 can be appropriately reflected as a cluster, and consequently the abnormality sign of the mechanical equipment 6 can be diagnosed appropriately and with high accuracy.

≪Modification≫
As described above, the abnormality sign diagnosis apparatuses 1 and 1A according to the present invention have been described by the respective embodiments. However, the present invention is not limited to these descriptions, and various modifications can be made.
For example, in the first embodiment, the “first sensor” used directly for the diagnosis of the abnormal sign of the internal combustion power generator 2 is the exhaust temperature sensor 32 (see FIG. 2), and the “second sensor” is the intake air temperature sensor. Although the case of 31 (see FIG. 2) and the generated power sensor 33 (see FIG. 2) has been described, the present invention is not limited thereto. That is, the cooling water temperature sensor that detects the temperature of the cooling water absorbed by the internal combustion engine 21 of the internal combustion power generator 2 is referred to as a “first temperature sensor”, and the intake air temperature sensor 31 and the generated power sensor 33 are referred to as a “second temperature sensor”. It is good.

Further, for example, a lubricating oil temperature sensor that detects the temperature of the lubricating oil absorbed by the internal combustion engine 21 of the internal combustion power generator 2 is referred to as a “first temperature sensor”, and the intake air temperature sensor 31 and the generated power sensor 33 are referred to as a “second temperature”. It may be a “sensor”.
In the first embodiment, the case where there are two types of “second sensors” (the intake air temperature sensor 31 and the generated power sensor 33) has been described, but the second sensor may be one type, There may be three or more types. The same applies to the second embodiment.

  Moreover, although 1st Embodiment demonstrated the case where the diagnostic object of an abnormal sign was the internal combustion power generator 2 (refer FIG. 2), it is not restricted to this. That is, “mechanical equipment” such as a gas turbine, a chemical plant, a distributed power generation facility, a medical facility, and a communication facility may be a diagnosis target for an abnormal sign. In this case, the “first sensor” and the “second sensor” included in the sensors installed in the “mechanical equipment” are selected in advance based on experiments and simulations.

In the second embodiment, the case where the cluster learning unit 181b (see FIG. 11) performs the learning process using the k-average method that is non-hierarchical clustering has been described, but the present invention is not limited to this. For example, the cluster learning unit 181b may perform learning processing based on fuzzy clustering, a mixed density distribution method, or the like as non-hierarchical clustering.
Moreover, although 2nd Embodiment demonstrated the case where the diagnostic object is the mechanical installation 6, in this mechanical installation 6, the internal combustion engine 21 (refer FIG. 2) and the generator 27 (refer FIG. 2) are included. The internal combustion power generator 2 (see FIG. 2) provided is also included. In this case, the “first sensor” includes the exhaust gas temperature sensor 32, and the “second sensor” includes the intake air temperature sensor 31 and the generated power sensor 33.

In each of the embodiments, the function deriving unit 14 (see FIGS. 1 and 10) has been described for obtaining a function representing the straight lines G _A and G _B (approximate curve: see FIG. 3). Absent. For example, a polynomial having an order of 2 or more may be used as the function representing the approximate curve, or an exponential function or a logarithmic function may be used.

  Each of the configurations shown in FIGS. 1 and 10 may be realized by hardware by designing a part or all of them with, for example, an integrated circuit. In addition, each of the above-described configurations may be realized by software by interpreting and executing a program that realizes each function by the processor. Information such as programs, tapes, and files that realize each function can be stored in a recording device such as a memory, a hard disk, or an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD. .

  Further, the control lines and information lines indicate what is considered necessary for the explanation, and not all the control lines and information lines on the product are necessarily shown. In practice, it may be considered that almost all the components are connected to each other.

DESCRIPTION OF SYMBOLS 1,1A Abnormal sign diagnostic apparatus 11 Communication means 12 Time series data acquisition means 13 Time series data storage means 14 Function derivation means 15 Function storage means (storage means)
16 Correction means 17 Corrected time series data storage means 18 Diagnosis means 18A Data mining means 181 Learning means 181a Learning object data acquisition part 181b Cluster learning part 181c Cluster information storage part 182 Diagnosis means 182a Diagnosis object data acquisition part 182b Abnormality measure calculation part 182c diagnostic unit 19 diagnostic result storage means 2 internal combustion power generator (mechanical equipment)
21 Internal combustion engine 27 Generator 31 Intake air temperature sensor (second sensor)
32 Exhaust temperature sensor (first sensor)
33 Power generation sensor (second sensor)
34 Communication means 6 Mechanical equipment

Claims (5)

Time-series data acquisition means for acquiring time-series data including a detection value of a first sensor installed in a mechanical facility and detection values of a plurality of types of second sensors installed in the mechanical facility;
Based on past time-series data in which the mechanical equipment is known to be normal, the detection value of the first sensor, the detection value of the second sensor correlated with the detection value of the first sensor, Storage means for storing a function representing an approximate curve of the distribution of the points specified by the plurality of types of the second sensors;
Correction means for correcting time-series data based on the function stored in the storage means;
Diagnostic means for diagnosing the presence or absence of abnormal signs of the mechanical equipment,
The correction means calculates a second difference by subtracting a predetermined reference value from a detection value of the second sensor, and calculates a first difference corresponding to the second difference with respect to the first sensor based on the function. The calculation process is executed for each of the plurality of types of the second sensors, and the first difference integrated value, which is the sum of the first differences, is subtracted from the detection value of the first sensor, thereby correcting the first sensor. Calculate the value,
The diagnostic means diagnoses that there is a sign of abnormality in the mechanical equipment when the correction value is out of a predetermined range ,
The correction unit performs a process of calculating the correction value related to the first sensor based on a plurality of types of detection values of the second sensor and the function as pre-processing of the diagnosis by the diagnosis unit. An abnormal sign diagnostic device. Time-series data acquisition means for acquiring time-series data including a detection value of a first sensor installed in a mechanical facility and detection values of a plurality of types of second sensors installed in the mechanical facility;
Based on past time-series data in which the mechanical equipment is known to be normal, the detection value of the first sensor, the detection value of the second sensor correlated with the detection value of the first sensor, Storage means for storing a function representing an approximate curve of the distribution of the points specified by the plurality of types of the second sensors;
Correction means for correcting time-series data based on the function stored in the storage means;
Learning means for learning a cluster indicating a range of a normal state of the mechanical equipment based on time-series data acquired during a period in which the mechanical equipment is known to be normal and further corrected by the correction means; ,
Diagnostic means for diagnosing the presence or absence of abnormal signs of the mechanical equipment,
The correction means calculates a second difference by subtracting a predetermined reference value from a detection value of the second sensor, and calculates a first difference corresponding to the second difference with respect to the first sensor based on the function. The calculation process is executed for each of the plurality of types of the second sensors, and the first difference integrated value, which is the sum of the first differences, is subtracted from the detection value of the first sensor, thereby correcting the first sensor. Calculate the value,
When the time series data including the correction value is outside the area of the cluster learned by the learning means, the diagnosis means diagnoses that there is a sign of abnormality in the mechanical equipment ,
The correction unit performs a process of calculating the correction value related to the first sensor based on a plurality of types of detection values of the second sensor and the function as pre-processing of the diagnosis by the diagnosis unit. An abnormal sign diagnostic device. The mechanical equipment is an internal combustion power generator comprising an internal combustion engine and a generator that generates electric power by driving the internal combustion engine,
The first sensor is an exhaust temperature sensor that detects a temperature of gas discharged from a combustion chamber of the internal combustion engine,
The second sensor includes
An intake air temperature sensor for detecting the temperature of air supplied to the combustion chamber;
The abnormality sign diagnosis apparatus according to claim 1, further comprising: a generated power sensor that detects generated power of the generator. A time-series data acquisition step of acquiring time-series data including a detection value of a first sensor installed in a mechanical facility and detection values of a plurality of types of second sensors installed in the mechanical facility;
Based on past time-series data in which the mechanical equipment is known to be normal, the detection value of the first sensor, the detection value of the second sensor correlated with the detection value of the first sensor, A storage step of storing a function representing an approximate curve of the distribution of points specified by the above in association with a plurality of types of the second sensors;
A correction step of correcting time series data based on the function stored in the storage step;
Diagnosing the presence or absence of abnormal signs of the mechanical equipment, and
In the correction step, a second difference is calculated by subtracting a predetermined reference value from a detection value of the second sensor, and the first difference corresponding to the second difference with respect to the first sensor is calculated based on the function. The calculation process is executed for each of the plurality of types of the second sensors, and the first difference integrated value, which is the sum of the first differences, is subtracted from the detection value of the first sensor, thereby correcting the first sensor. Calculate the value,
In the diagnosis step, when the correction value is out of a predetermined range, the machine equipment is diagnosed as having an abnormal sign ,
In the correction step, as a pre-process of the diagnosis step, a process of calculating the correction value for the first sensor based on a plurality of types of detection values of the second sensor and the function is executed. Abnormal sign diagnosis method. A time-series data acquisition step of acquiring time-series data including a detection value of a first sensor installed in a mechanical facility and detection values of a plurality of types of second sensors installed in the mechanical facility;
Based on past time-series data in which the mechanical equipment is known to be normal, the detection value of the first sensor, the detection value of the second sensor correlated with the detection value of the first sensor, A storage step of storing a function representing an approximate curve of the distribution of points specified by the above in association with a plurality of types of the second sensors;
A correction step of correcting time series data based on the function stored in the storage step;
A learning step of learning a cluster indicating a range of a normal state of the mechanical equipment, based on time-series data acquired during a period in which the mechanical equipment is known to be normal and further corrected by the correction step; ,
Diagnosing the presence or absence of abnormal signs of the mechanical equipment, and
In the correction step, a second difference is calculated by subtracting a predetermined reference value from a detection value of the second sensor, and the first difference corresponding to the second difference with respect to the first sensor is calculated based on the function. The calculation process is executed for each of the plurality of types of the second sensors, and the first difference integrated value, which is the sum of the first differences, is subtracted from the detection value of the first sensor, thereby correcting the first sensor. Calculate the value,
In the diagnosis step, when the time series data including the correction value exists outside the area of the cluster learned in the learning step, the machine facility is diagnosed as having an abnormal sign ,
In the correction step, as a pre-process of the diagnosis step, a process of calculating the correction value for the first sensor based on a plurality of types of detection values of the second sensor and the function is executed. Abnormal sign diagnosis method.

Images