JP6503541B2 - Pump abnormality detection system, pump abnormality detection method, and pump abnormality detection program - Google Patents

Description

  The present invention relates to sensing technology for detecting an abnormality in a pump that pumps fluid, and more particularly to a pump abnormality detection system for detecting an abnormality in a pump used in a power plant, a pump abnormality detection method, and a pump abnormality detection program.

  Various pumps used in power plants have been post-mortemly analyzed (such as frequency analysis) on the waveform of a vibration sensor attached to the pump by an information processing apparatus in order to identify abnormal factors. In addition, in constant surveillance, surveillance was conducted by a hearing rod and visual observation. In addition, confirmation methods that rely on the past experience and expertise of the supervisors are also being conducted.

  In analysis by the information processing device, by using the engineer's experience together with pump design values such as the inner diameter of the bearing and the diameter of the ball bearing, the vibration period and frequency component are specified, for example, identification of abnormal vibration location or bearing flaw etc. The factor of On the other hand, as in the analysis of one specific vibration sensor, the object for verifying the abnormality was artificially narrowed down, and the analysis of the abnormality factor, the failure prediction, etc. was performed for the object.

  In addition, the technique relevant to abnormality monitoring of a pump is described in patent document 1. FIG. This document describes a device for providing a vibration sensor inside the pump and automatically stopping the operation of the pump when a clear abnormality is detected by the vibration sensor.

Japanese Utility Model Publication No. 02-145696

  Although recent power plant systems also require automatic shutdown of the system, maintenance of the operation status is also important, and in the case of an abnormality that does not cause any problems in operation, the system will not be shut down.

  In addition, if it is possible to carry out abnormal detection that leads to detection of a minor abnormality or failure prediction, it works beneficially for operation.

  On the other hand, as exemplified above, when analysis target is determined by artificially important pump and frequency analysis etc. are performed on measured values obtained from sensors attached to specific elements, design information Various advance preparation and analysis experts based on installation information are needed. As a result, the larger the system, the more difficult it is to perform anomalous verification and prediction comprehensively in the entire system.

  In addition, the listening stick and visual monitoring are human subjective evaluations, and quantitative evaluation is difficult. In addition, it is difficult to perform comprehensive constant monitoring of all devices. In addition, some power plant systems have pumps installed at locations where human access is difficult.

  Therefore, an object of the present invention is to provide a pump abnormality detection system capable of detecting an abnormality occurring in a pump in a power plant with good convenience.

The pump abnormality detection system according to the present invention relates to all of the individual vibration sensor means at each point in time with respect to a plurality of vibration sensor means for constantly measuring and measuring the vibration of one pump for delivering the fluid provided in the power plant. Model storage means for storing a model of normal operation data in which no abnormality has occurred in the pump, showing mutual relationships among combinations of output values, and measurement data measured by the plurality of vibration sensor means and the sensor information storage means for collecting and storing respectively, when the acquiring the normal operation data, and adjusting means for varying the pressure of fluid flowing through the pump gradually accumulated by the sensor information storage means, of the plurality of vibration sensor means a pressure of the fluid obtained by changing gradually flowing to the pump abnormality in the pump has not occurred From the normal operation data including constant data, and model construction unit which extracts a correlation between each of the measured data, to build the model of the normal operation data is recorded in the model storing unit, said model The correlation between the correlation obtained by the model recorded in the storage means and the combination of all the output values at each time point included in the measurement data of any period obtained from the plurality of vibration sensor means is shown. Pump abnormality detection means for comparing occurrences of correlations with each other to monitor occurrence of abnormal values of the amount of collapse based on the amount of collapse of the correlation, and extracting relationship collapse of a predetermined amount as a notification event related to the pump , And.

A pump abnormality detection method according to the present invention comprises a plurality of vibration sensor means for constantly measuring vibration of one pump for sending out a fluid provided in a power plant, and adjusting means for changing the pressure of the fluid flowing to the pump The measurement data measured by the vibration sensor when the pressure of the fluid flowing to the pump is gradually changed by the adjustment unit is collected and accumulated as normal operation data. The plurality of vibrations obtained by collecting and accumulating measurement data of an arbitrary period measured by the sensor means and gradually changing the pressure of the fluid flowing to the pump in which no abnormality has occurred in the accumulated pump. from the normal operation data including the measurement data of the sensor means, extracting a correlation of all combinations of output values of the measured data of each of the time points Te, wherein recorded in the model storing means to build a model of the normal operation data, the mutual relationship existing between the combination of all the output values of the time included in the arbitrary period and correlation obtained by said model The correlation between the measured data shown is compared to monitor occurrence of an abnormal value of the amount of collapse based on the amount of collapse of the correlation, and extract a predetermined amount of relationship collapse as a notification event related to the pump It is characterized by

The pump abnormality detection program according to the present invention is a program for detecting a plurality of vibration sensor means for constantly setting and measuring the vibration of a pump provided in the power plant for controlling the control unit of the information processing apparatus. A plurality of combinations obtained by gradually changing the pressure of the fluid flowing to the pump in which no abnormality has occurred in the pump, showing the interrelationship among the combinations of all the output values of the vibration sensor means of Correlation obtained by the model stored in the model storage unit storing the model of the normal operation data including the measurement data of the vibration sensor unit, and each time point included in an arbitrary period obtained from the plurality of vibration sensor units collapse for comparing processing and correlation relative to the collapse of the correlation between the measurement data showing the mutual relationship existing between the combination of all the output values of Monitoring the occurrence of abnormal values, the pump failure detecting means for extracting the relationships collapse a predetermined amount as a notification event related to the pump, when acquiring the normal operation data, bit by bit the pressure of fluid flowing through the pump while exercising adjusting means for changing measured by the plurality of vibration sensor means, from said normal operation data abnormality in the pump has not occurred, to extract the correlation between each of the measured data, the normal It is characterized in that it operates as model construction means for constructing a model of hour movement data and recording it in the model storage means.

  According to the present invention, it is possible to provide a pump abnormality detection system capable of detecting an abnormality occurring in a pump in a power plant with good convenience.

1 is a block diagram showing a pump abnormality detection system according to a first embodiment of the present invention. It is a flowchart which illustrates the processing operation of the pump abnormality detection system concerning a 1st embodiment. It is a block diagram which shows the pump abnormality detection system concerning the 2nd Embodiment of this invention. BRIEF DESCRIPTION OF THE DRAWINGS It is explanatory drawing which shows the pump abnormality detection system concerning an Example. It is the flowchart which illustrated the processing operation of the sensor information storage part. It is the flowchart which illustrated the processing operation of the model construction part. It is the flowchart which illustrated processing operation of the pump abnormality detection section.

  Hereinafter, a pump abnormality detection system according to an embodiment of the present invention will be described based on the drawings.

First Embodiment
FIG. 1 is a block diagram showing a pump abnormality detection system 10 according to a first embodiment.

  The pump abnormality detection system 10 includes a pump abnormality detection unit 20 and a model storage unit 30 for receiving as an input a plurality of vibration sensor means for measuring the vibration of the pump provided in the power plant and detecting an abnormality. Involve and configure.

  The model storage unit 30 stores a normal model showing the relationship between measurement data between output values of an arbitrary period for each vibration sensor (1-1, 1-2,... 1-n). It is done. The normal state model may be generated from operation data of a vibration sensor group that does not assume an abnormality operating the pump as one method.

  The pump abnormality detection unit 20 obtains a correlation between measurement data of measurement data of a plurality of vibration sensor means to be monitored, such as measurement values having real-time property and accumulated measurement values. , Compare with the correlation found in the normal mode model recorded in the model storage unit 30. The pump abnormality detection unit 20 detects a predetermined amount of loss of relationship based on the amount of loss of correlation (the amount of mismatch) in the comparison result, and extracts it as a notification event when an abnormal value occurs. Further, together with the amount of collapse, the occurrence of abnormality may be extracted on the basis of the manner of collapse. In addition, it is possible to take account of the change (change amount) per hour of the amount of collapse. The extracted notification event may be recorded in a predetermined storage unit, automatically notified to the person in charge, or notified to another system. Also, depending on the degree of abnormality, a signal for automatically stopping the pump may be output. When the amount of collapse or change per hour exceeds a predetermined value, notification may be given as a notification event having an emergency degree.

Next, an operation example of the pump abnormality detection system 10 according to the first embodiment will be described.
FIG. 2 is a flowchart illustrating the processing operation of the pump abnormality detection system 10 according to the first embodiment.

  The pump abnormality detection unit 20 receives measurement data of the vibration sensor group to be monitored (S101). The measurement data input at this time may be real-time measurement data or measurement data collected in the past of an arbitrary section.

  In parallel, the pump abnormality detection unit 20 acquires a normal model recorded in the model storage unit 30 (S102).

  The pump abnormality detection unit 20 identifies the received sensor value group and obtains a correlation (S103).

  Next, the pump abnormality detection unit 20 compares the found correlation with the correlation found in the normal state model to calculate a value indicating a loss of relationship, and whether or not the relationship has a loss It is determined where it is, where it is located, when it is located, etc. (S104).

The pump failure detection unit 20 notifies a predetermined device as a notification event when the relationship collapse / progression of relationship / progressive collapse at a predetermined position or the like is a predetermined amount or more (S105).
As described above, according to the present embodiment, it is possible to obtain a detection system that can easily and accurately catch an abnormality of a corresponding pump from measurement values of a plurality of vibration sensors installed in the pump. At this time, as in the conventional system, it is possible to determine an abnormality without requiring determination by an analysis expert. In addition, it is possible to build a real-time monitoring system with low resources.

  In the present system, analysis may be performed for each pump, and a plurality of pumps can be simultaneously monitored in parallel.

Second Embodiment
Next, a second embodiment will be described. In addition, description is simplified or abbreviate | omitted about the location similar to 1st Embodiment.
The second embodiment includes means for constructing a normal state model in which no abnormality occurs in the pump used in the first embodiment.

  FIG. 3 is a block diagram showing the control rod monitoring system 10 of the second embodiment.

  The pump abnormality detection system 10 is configured to include a pump abnormality detection unit 20, a model storage unit 30, a sensor information storage unit 40, and a model construction unit 50.

  The sensor information accumulation unit 40 collects and accumulates measurement data measured by a plurality of vibration sensors associated with each pump.

The model construction unit 50 is configured to measure the operation data groups stored in the sensor information storage unit 40 during normal times in which no abnormality has occurred in one or all of the pumps. The correlation is extracted, and a normal model of each pump of normal operation data is constructed and recorded in the model storage unit 30.
The model construction unit 50 may process the model normal model of each pump by generating an approximate expression and a fit value indicating the correlation between measurement data measured by the vibration sensor group for each pump. The generated approximate expression and fit value are recorded in the model storage unit 30 as a model of each pump. Further, it is more preferable that the model constructing unit 50 has an algorithm that compares the generated fit value with a predetermined threshold and adopts it as a normal model when the fit value is equal to or more than the predetermined threshold.

  By generating a model using an approximate expression in this manner, processing resources can be reduced and real-time processing performance can be improved.

As described above, when using the normal-time model using the approximate expression, the pump abnormality detection unit 20 may be operated for each pump as follows.
The pump abnormality detection unit 20, for example, converts one of the measurement data measured by an arbitrary vibration sensor (for example, a sensor in the vicinity of a key element such as a bearing) around each pump into an approximate expression of a normal model of the corresponding pump. substitute. By this substitution, each predicted value to be measured by the vibration sensor can be calculated.
Next, the pump failure detection unit 20 determines the amount of loss of the correlation by comparing the quantity of the difference between the predicted value and the actual value with a predetermined threshold. An abnormality is determined when the amount of collapse of the correlation exceeds a predetermined threshold.

  As described above, according to the present embodiment, it is possible to obtain a detection system that can easily and accurately catch an abnormality of a corresponding pump from measurement values of a plurality of vibration sensors installed in the pump. At this time, as in the conventional system, it is possible to determine an abnormality without requiring determination by an analysis expert. In addition, it is possible to build a real-time monitoring system with low resources.

The invention will now be described with reference to one example. The present embodiment is a system for monitoring, in real time, slight scratches and grease consumption in bearings that may occur in one pump in the second embodiment. Also, a plurality of vibration sensors are respectively disposed in the vicinity of a bearing with the motor. In addition, it is good also as providing in a pump.
In other words, the system of this embodiment operates as a bearing flaw detection device.

  As a part of the power generation facility, as shown in FIG. 4, a pump 2 for delivering the fluid flowing in the pipe, a motor 3 for operating the pump 2, and a regulator 4 for regulating the fluid are provided.

  The pump abnormality detection system (bearing scratch detection device) monitors the pump 2. Vibration sensors 1-1 and 1-2 are attached to the pump 2. Each vibration sensor is applied to measure the vibration in the vicinity of the bearing. The present embodiment will be described by using at least two vibration sensors. Three or more vibration sensors may be installed in the pump 2. In addition, other types of sensors (acoustic sensor, temperature sensor, etc.) may be provided together. The adjuster 4 adjusts the amount, pressure, and the like of the fluid in the pipe 1. In the present embodiment, the regulator 4 is described as a control valve that controls the water pressure in the pipe.

  The pump abnormality detection system 10 receives an output value of the vibration sensors 1-1 and 1-2 as an input to detect an abnormality. In addition, the output of the adjustment machine 4 is also used together with the output values of the vibration sensors 1-1 and 1-2 to construct a normal-time model.

  The sensor information storage unit 40 associates the output values of the vibration sensors 1-1 and 1-2 with the time data of the measurement time and sequentially stores them.

  The model construction unit 50 acquires measurement data for a certain period of time during normal operation with no bearing damage from the sensor information storage unit 40, extracts the correlation of the measurement data together with the output value of the adjuster 4, and is normal Build a time model.

  The pump abnormality detection unit 20 operates as a bearing flaw detection unit. The pump abnormality detection unit 20 receives the extracted correlation from the model storage unit 30, receives measurement data for a certain period of time from the sensor information storage unit 40, and detects the occurrence of a bearing flaw in the pump 2 based on these. Do.

  The pump abnormality detection unit 20 includes, as necessary, a notification unit that notifies the device as required, such as an alarm device or a display, and notifies the occurrence of the detected bearing damage by an alarm sound, a message, or the like. Further, for example, a blemish less than a predetermined value may only be logged.

Next, operations of the vibration sensors 1-1 and 1-2 and the sensor information storage unit 40 will be described using FIG.
First, the vibration sensors 1-1 and 1-2 constantly measure the vibration of the pump 2 (step S 501).
Next, the sensor information storage unit 40 receives each measurement data measured by the vibration sensors 1-1 and 1-2 together with time data of measurement time (step S 502).
Next, the sensor information storage unit 40 stores data (measurement data and time data) received from the vibration sensors 1-1, 1-2 (step S503).
The operations of steps S501 to S503 described above are constantly and repeatedly performed.
The accumulated information is composed of output values from the vibration sensors 1-1 and 1-2 and measurement time, and is in a form generally called time-series data.
As an aspect of information accumulation by the sensor information accumulation unit 40, a mechanism such as a relational database may be used, or a simple text file may be held.

Next, the operation of the model construction unit 50 will be described using FIG.
First, at the time of normal operation in which no bearing damage occurs in the pump 2, the regulator 4 is adjusted to gradually change the water pressure flowing to the pump 2 (step S601). At this time, the direction of raising and lowering the water pressure may be mixed.

Next, the vibration sensors 1-1 and 1-2 read the vibration due to the water pressure that changes due to the operation of the adjuster 4, and the sensor data storage unit 40 stores the measured data and time data as needed (step S 602, FIG. Steps S501 to S503).
Next, the model construction unit 50 receives, from the sensor information storage unit 40, measurement data in a period in which the water pressure is changed by operating the adjuster 4 (a period in which fluctuation data during normal operation is recorded) (step S603). ).
Next, the model construction unit 50 determines whether there is a correlation between the vibration values measured for the pump 2 from the received measurement data (step S604).

  Here, from the time series data of a fixed time of two measured points obtained from the sensor information storage unit 40, the model constructing unit 50 approximates a formula-based approximation such as B = f (A) as a correlation between the two points. Generate an expression As a method of generating an approximate expression, for example, a method called linear regression may be used. In addition, various other methods have been proposed, and any method may be adopted.

  Furthermore, the model construction unit 50 generates a fit value, which is an index of how much the approximate expression approximates the actual data, from the generated approximate expression and the time-series data used at the time of generation. When approximating using the least squares method as linear regression, the fitted value can be the determination coefficient in the least squares method.

  Next, the fit value is compared with a predetermined threshold value, and if it is equal to or more than the threshold value (N in step S605), the relationship between the two points (approximate expression and fit value) is stored as a model, and the process is ended ( Step S606). If the fit value is less than or equal to the threshold (Y in step S605), the process ends. When three or more vibration sensors are used, a model may be generated using the relationship between multiple points. Hereinafter, the relationship between the stored two normal points is called a model.

Next, the operation of the pump abnormality detection unit 20 for detecting a bearing damage will be described with reference to FIG.
It is assumed that a model at normal time is constructed as the operating condition of the pump abnormality detection unit 20, and measurement data from the vibration sensors 1-1 and 1-2 are constantly stored in the sensor information storage unit 50. .

  First, the pump abnormality detection unit 20 acquires, from the sensor information storage unit 50, measurement data for a certain past time from a certain time t at which it is desired to detect a bearing flaw (step S701). Here, a certain time t is a slightly past time than the current time. If measurement data of the current time is always accumulated in the sensor information storage unit 50, the time t may be the current time.

  Next, a normal model stored in the model construction unit 50 is acquired (step S702).

  Next, the relationship between the vibration sensors 1-1 and 1-2 at two points A and B (approximate expression B = f (A) and fit value) is acquired from the model (step S 703).

  Next, the pump abnormality detection unit 20 substitutes the value of one vibration sensor A included in the measurement data obtained from the sensor information storage unit 50 into the approximate expression B = f (A), and the other vibration sensor as a result The predicted value of B is obtained (step S704).

  Next, the pump abnormality detection unit 20 calculates a difference R between the obtained predicted value of the vibration sensor B and the actually measured value of the vibration sensor B obtained from the sensor information storage unit 40 (step S705).

  Next, when the difference R exceeds the predetermined threshold (Y in step S706), the pump abnormality detection unit 20 determines that the relation of the approximate expression B = f (A) is not established, and the pump It is determined that there is a possibility that bearing damage has occurred in 2 and a predetermined notification action is executed (step S 707).

  Next, after the notification of step S707, the pump abnormality detection unit 20 advances time t by a fixed time Δt, including the case where the difference R does not exceed the predetermined threshold (step S706N). And repeat the process from step S701.

  Here, Δt is set from the interval at which it is desired to detect a bearing flaw, but it needs to be larger than the interval at which the vibration sensors 10 and 20 notify the sensor information storage unit 50 of the measurement result.

  In addition, the pump abnormality detection system 10 obtains the model bi-directionally in advance such as the approximate expression B = f (A), the approximate expression A = f (B), and each fit value in advance, and the vibration sensor A and the vibration sensor The reference vibration value of B may be switched and monitored sequentially.

  In the bearing flaw detection device according to the present embodiment obtained in this manner, the bearing in the pump 2 is measured using the measurement data of the existing vibration sensors 1-1 and 1-2 installed to measure the vibration. It is possible to detect that a wound has occurred.

  In addition, when using three or more measured values, there is also an event that can identify the propagation source of the collapse by bilaterally analyzing.

  In addition, correlations with various sensors disposed in the pump may be similarly analyzed. For example, a pressure sensor that measures the pressure at the input port of the pump and the pressure at the output port, or an audible sound sensor can also be used.

  Further, in the bearing flaw detection device of the present embodiment, it is possible to detect in the initial stage of bearing flaw occurrence. It is also possible to detect that the wear of the grease has exceeded a predetermined value.

  In addition, it is possible to detect an abnormality in any part of the bearing without consuming a large amount of computing resources. This detection also includes anticipation of the occurrence of the problem.

  It is very difficult for human beings to recognize early-stage wounds with a hearing rod or the like without using such analysis methods. In addition, it is necessary to make an inspection target as an abnormal point, and it is very difficult to perform all on equipment as resource allocation. On the other hand, monitoring of the value of each sensor as in the present method has the advantage of being able to centrally monitor various matters such as the identification of an abnormal point and the degree of abnormality in a wide range.

  Each part of the pump abnormality detection system may be realized using a combination of hardware and software. In the combination of hardware and software, the pump abnormality detection program is expanded in the RAM, and each unit is realized as various means by operating hardware such as a control unit (CPU) based on the program. Also, this program may be fixedly recorded on a recording medium and distributed. The program recorded on the recording medium is read into the memory via a wired, wireless, or recording medium itself to operate the control unit and the like. Incidentally, examples of the recording medium include an optical disk, a magnetic disk, a semiconductor memory device, a hard disk and the like.

  If the above embodiment is described in another expression, the information processing apparatus for operating as a pump abnormality detection system is a pump abnormality detection means, a model storage means, a sensor information storage means based on a pump abnormality detection program expanded in a RAM. This can be realized by operating the control unit as model construction means or the like.

  As described above, the information processing apparatus to which the present invention is applied can provide a pump abnormality detection system capable of detecting an abnormality occurring in a pump in a power plant with good convenience.

  Further, the specific configuration of the present invention is not limited to the above embodiments and examples, and there may be changes such as separation / merging of block configurations, replacement of procedures, and the like without departing from the scope of the present invention. The above description does not limit the present invention. For example, each component (a pump abnormality detection unit, a sensor information storage unit, and the like) may be configured by each information processing apparatus.

1 · · · Vibration sensor 2 · · · Pump 3 · · · Motor 4 · · · Adjustment machine (adjustment means)
10 · · · Pump malfunction detection system 20 · · · Pump malfunction detection unit (pump malfunction detection means)
30 · · · Model storage unit (model storage means)
40 · · · Sensor information storage unit (sensor information storage means)
50 · · · Model construction unit (model construction means)

Claims (8)

With respect to a plurality of vibration sensor means for permanently measuring and measuring the vibration of one pump for delivering the fluid provided in the power plant, there is a mutual relationship among combinations of all output values of the individual vibration sensor means at each time. Model storage means for storing a model of normal operation data in which no abnormality has occurred in the pump;
Sensor information storage means for collecting and storing measurement data measured by the plurality of vibration sensor means;
When the acquiring the normal operation data, and adjusting means for varying the pressure of fluid flowing through the pump gradually,
The normal operation including measurement data of the plurality of vibration sensor means acquired by gradually changing the pressure of the fluid flowing to the pump in which no abnormality has occurred in the pump accumulated by the sensor information accumulation means. Model construction means for extracting correlations between respective measurement data from data, constructing a model of the normal operation data, and recording the model in the model storage means;
The correlation found between the model stored in the model storage means and the correlation between the combinations of all the output values at each time point included in the measurement data of an arbitrary period obtained from the plurality of vibration sensor means And the occurrence of an abnormal value of the amount of collapse based on the amount of collapse of the correlation, and a pump abnormality detection of extracting a predetermined amount of relationship collapse as a notification event related to the pump Means,
A pump abnormality detection system comprising: The model construction means generates an approximate expression and a fit value indicating a correlation between measurement data of the normal operation data measured by the vibration sensor, and stores the approximate expression and the fit value as a model. The pump abnormality detection system according to claim 1, characterized in that The model construction unit, the generated the fit value is compared with a predetermined threshold, and if the fit value is a predetermined threshold or more, claim you and to store the model The pump abnormality detection system according to 2.   The pump abnormality detection system according to any one of claims 1 to 3, further comprising: notification means for notifying a predetermined object of the fact when an abnormality is detected by the pump abnormality detection means.   The pump abnormality detection system according to any one of claims 1 to 4, wherein the pump is a pump that delivers a liquid.   The pump abnormality detection means substitutes one of the measurement data measured by the vibration sensor means into the approximate expression generated by the model construction means, thereby other than the vibration sensor means used for substitution. A predicted value of one or a plurality of measured data is calculated, and the predicted value is compared with an actual measured value of each of the plurality of vibration sensor means, and the amount of the difference exceeds a predetermined threshold value. The pump abnormality detection system according to claim 4 or 5, wherein the amount of breakdown of the relationship is determined. A plurality of vibration sensor means for constantly measuring and measuring the vibration of one pump for sending out the fluid provided in the power plant, and adjustment means for changing the pressure of the fluid flowing to the pump are provided as normal operation data Measurement data measured by the vibration sensor means when the pressure of the fluid flowing to the pump is gradually changed by the adjusting means is collected and accumulated, and measurement of any period measured by the vibration sensor means Collect and accumulate data,
The normal operation data including the measurement data of the plurality of vibration sensor means acquired by gradually changing the pressure of the fluid flowing to the pump in which no abnormality has occurred in the pump which has already been stored of extracting a correlation of all combinations of output values of the measurement data, recorded in the model storing unit to build the model of the normal operation data,
The amount of loss of correlation is compared by comparing the correlation found by the model and the correlation between measured data showing the mutual relationship among combinations of all output values at each time point included in the arbitrary period. A pump abnormality detection method, comprising: monitoring occurrence of an abnormal value of the amount of collapse based on and extracting a predetermined amount of relationship collapse as a notification event related to the pump. The control unit of the information processing apparatus
With respect to a plurality of vibration sensor means for permanently measuring and measuring the vibration of one pump for delivering the fluid provided in the power plant, there is a mutual relationship among combinations of all output values of the individual vibration sensor means at each time. A model of normal operation data including measurement data of the plurality of vibration sensor means acquired by gradually changing the pressure of the fluid flowing to the pump without abnormality in the pump, A measurement showing the correlation found by the model stored in the model storage means and the correlation between the combinations of all the output values at each time point included in an arbitrary period obtained from the plurality of vibration sensor means by comparison a correlation between the data, and monitors the occurrence of outliers collapse amount relative to collapse of correlation relates the pump relationships collapse a predetermined amount A pump abnormality detecting means for extracting as a governor elephant,
When acquiring the normal operation data, the pressure of fluid flowing through the pump is measured by said plurality of vibration sensor means while exercising adjusting means for changing little by little, the normal no abnormality occurred in the pump Model construction means for extracting correlations between respective measurement data from operation data, constructing a model of the normal operation data, and recording the model in the model storage means;
The pump abnormality detection program characterized by operating as.

Images