State monitoring system

Abstract

Provided is a state monitoring system which can be installed as a low-cost, small-sized and power-saving device having limited computing resources, and which can monitor and diagnose the state of a device to be monitored with a relatively small amount of computation. A state monitoring system (10) is provided with a measurement data processing unit that calculates processed data from measurement data, and the measurement data processing unit (11) executes: a fine division step in which time-series data along a time series, which is the measurement data, is finely divided in a time direction; an FFT step in which spectrum data is calculated by high-speed Fourier conversion for each of the finely divided data obtained by finely dividing the time-series data; a statistical step in which a partial spectrum of a partial frequency band is extracted from the spectrum data and a statistical magnitude of the partial spectrum is calculated; and an integration step: in the integration step, arranging the statistical magnitude according to the time sequence of the original fine segmentation data, and taking the statistical magnitude as the processed data.

Description

Technical Field

[0001] The present invention relates to a system for monitoring the status of a mechanical component or the like. Background Art

[0002] To detect abnormalities in machinery or perform diagnostics, sound and vibration are often measured and analyzed.

[0003] Patent document 1 proposes an invention as an abnormality diagnosis device for mechanical equipment, which uses the frequency components of diagnostic spectrum data based on frequency analysis performed through envelope analysis and high-speed Fourier transform (hereinafter sometimes abbreviated as "FFT") for the measured waveform to determine the presence or absence of abnormalities and the location of abnormalities.

[0004] Patent Document 2 proposes a method for diagnosing rolling bearing damage. This method acquires the bearing's vibration acceleration in a time series format, performs bandpass filtering and envelope processing in the analog domain, and then performs A / D (analog / digital) conversion. The converted data is then subjected to an FFT. The peak frequency of the FFT result is compared with the characteristic frequency to determine rolling bearing damage.

[0005] Patent Document 1: Japanese Patent Application Laid-Open No. 2007-285875

[0006] Patent Document 2: Japanese Patent Application Laid-Open No. 9-113416

[0007] However, if the objects being inspected and diagnosed are numerous or small, the computing resources allocated to monitoring and diagnosing their conditions are limited. If the diagnostic equipment is also required to be inexpensive, compact, and power-efficient, computationally intensive monitoring and diagnostic methods are difficult to implement.

[0008] In the abnormality diagnosis device described in Patent Document 1, the number of waveform data items measured is large, and therefore enormous computational resources are required for filter processing accompanying envelope processing and FFT processing for frequency analysis.

[0009] In the damage diagnosis method described in Patent Document 2, the bandpass filtering and envelope processing performed in the analog domain require a large number of operational amplifiers and high-precision passive components. This increases the cost of the device executing the method and is prone to noise superposition. Furthermore, power consumption is high. Furthermore, FFT processing tends to require more computing resources. Summary of the Invention

[0010] In view of the above background, the present invention aims to provide a condition monitoring system that can be implemented as an inexpensive, compact, and power-saving device with limited computing resources and can monitor and diagnose the condition of a monitored device with relatively little computing power.

[0011] In order to solve the above-mentioned problems, the present invention adopts the following first structure, that is, a condition monitoring system including a measurement data processing unit that calculates processed data based on measurement data, wherein:

[0012] The measurement data processing unit performs the following:

[0013] a fine segmentation step of finely segmenting the time series data along the time series, which is the measurement data or the first processed data calculated based on the measurement data, in a time direction;

[0014] An FFT step, in which spectrum data is calculated by fast Fourier transform for each subdivided data after the time series data is subdivided;

[0015] a statistical step, in which a partial spectrum of a certain frequency band is extracted from the spectrum data and a statistic of the partial spectrum is calculated; and

[0016] In the integration step, the statistics are arranged according to the time series of the original finely segmented data as the processed data.

[0017] That is, the subdivided data, after being finely segmented in the time direction, is subjected to calculation of spectral data, extraction of a portion of the frequency band, and calculation of statistics. Finally, the statistics are arranged in a time series. Because the high-load processing can be subdivided, the number of processes can be reduced, resulting in processed data that requires fewer computing resources than directly using the time series data. This subdivision is performed at least 10 times, preferably 100 times or more. Furthermore, since specific frequency bands are extracted from the spectrum of the subdivided data, it is necessary to ensure that the number of data points in each subdivided data point is sufficient to prevent the frequency resolution of the spectrum from being too low. Specifically, the number of data points in each subdivided data point is ensured to be at least 10, preferably 100 or more.

[0018] The state monitoring system according to the present invention can adopt the following second structure, that is,

[0019] As the measurement data, time-series data of any one of the physical quantities of the monitored device, such as jerk, angular jerk, acceleration, angular acceleration, displacement, angular displacement, position, phase, and pressure, is used.

[0020] The state monitoring system according to the present invention can adopt the following third structure, that is,

[0021] In addition to the above structure, a diagnosis unit is further provided, which performs:

[0022] A second FFT step, in which second spectrum data is calculated based on the processed data by high-speed Fourier transform; and

[0023] The judgment step comprises calculating a judgment result as a statistical value or a judgment value based on the second spectrum data.

[0024] Furthermore, the state monitoring system according to the present invention can adopt the following fourth configuration:

[0025] In any one of the first to third configurations, the partial frequency band for which the extraction is performed can be changed, and a plurality of processed data having different partial frequency bands can be calculated.

[0026] Furthermore, the state monitoring system according to the present invention can adopt the following fifth configuration:

[0027] In any one of the first to fourth structures, in the fine division step, the number of data of the finely divided data is set to a preset number of data, or the number of subdivisions of the finely divided data is set to a preset number of divisions.

[0028] Furthermore, the state monitoring system according to the present invention can adopt the following sixth configuration:

[0029] In any one of the first to fifth configurations, the fine division step is started before the measurement data processing unit acquires all of the measurement data.

[0030] Furthermore, the state monitoring system according to the present invention can adopt the seventh configuration as follows:

[0031] In any one of the first to fifth configurations, the fine division step is started after the measurement data processing unit acquires all of the measurement data.

[0032] Furthermore, the state monitoring system according to the present invention can adopt the following eighth configuration:

[0033] In any one of the first to seventh configurations, each of the subdivided data is divided so as to overlap with other subdivided data adjacent thereto in time series by a predetermined ratio.

[0034] Furthermore, the state monitoring system according to the present invention can adopt the following ninth configuration, namely,

[0035] In any one of the first to eighth configurations, the spectrum data is any one of an amplitude spectrum, a power spectrum, and a power density spectrum.

[0036] Furthermore, the state monitoring system according to the present invention can adopt the following tenth configuration:

[0037] In any one of the first to ninth configurations, the statistic is any one of a maximum value, a total value, an average value, and a root mean square value of the partial spectrum.

[0038] Furthermore, the state monitoring system according to the present invention can adopt the following eleventh structure, that is,

[0039] In any one of the third to tenth configurations and referring to the third configuration, the statistical value in the determination step of the diagnosis unit is any one of a maximum value, a total value, an average value, and a root mean square of the second spectrum data.

[0040] The condition monitoring system of the present invention processes large amounts of time-series data, not directly, but by dividing it into sub-segments along the time axis. This reduces the memory and computational complexity required for each calculation, enabling it to be installed even on inexpensive, compact, and power-efficient devices with limited computing resources. This makes it possible to monitor the condition of various devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a diagram showing an example of a basic configuration of a condition monitoring system according to the present invention.

[0042] Figure 2 This is a diagram illustrating a first flow example of how the condition monitoring system according to the present invention calculates processed data from measurement data.

[0043] Figure 3 This is a graph showing an example of time series data processed by the condition monitoring system according to the present invention.

[0044] Figure 4 This example shows how spectrum data, partial spectra, and statistical quantities are calculated from time series data.

[0045] Figure 5 This is a diagram showing a second flow example of calculation of processed data for a plurality of frequency bands by the condition monitoring system according to the present invention.

[0046] Figure 6 This is a diagram of a third flow example of calculation of processed data for a plurality of frequency bands by the condition monitoring system according to the present invention.

[0047] Figure 7 3 is a graph showing an example of second spectrum data obtained by performing fast Fourier transform on processed data.

[0048] Figure 8 It means according to Figure 7 A graph of the maximum value and RMS calculated for the second spectrum data.

[0049] Figure 9 It means from Figure 7 A graph showing the root mean square calculated by extracting the range before and after the peak of the second spectrum data.

[0050] Figure 10A Yes Figure 7 FIG2 is a diagram showing an example of a spectrum after weighting of the second spectrum data.

[0051] Figure 10B Yes Figure 10A Graph of the function used for weighting in .

[0052] Figure 11 is a graph showing an example of processed data.

[0053] Figure 12 Yes Figure 11 Figure 1 is an example of the processed data after bandpass filtering.

[0054] Figure 13 Yes Figure 11 Figure 1 shows an example of the processed data after squaring.

[0055] Figure 14 Yes Figure 11 Figure 1 shows an example of processed data after subtracting the noise level.

[0056] Figure 15 This is a functional block diagram showing a first configuration example when the condition monitoring system according to the present invention is implemented.

[0057] Figure 16 This is a functional block diagram showing a second configuration example when the condition monitoring system according to the present invention is implemented.

[0058] Figure 17 This is a functional block diagram showing an application mode in which a display device is added to the second structural example.

[0059] Figure 18 This is a functional block diagram showing an application method in which a server and a display device are added to the second structural example.

[0060] Figure 19 This is a functional block diagram showing a third configuration example when the condition monitoring system according to the present invention is implemented.

[0061] Figure 20 This is a functional block diagram showing an application method in which a server is added to the field network of the third configuration example. DETAILED DESCRIPTION

[0062] The present invention provides a condition monitoring system for monitoring the condition of monitored equipment, such as machinery and components. The condition monitoring system includes a measurement data processing unit that calculates processed data based on measurement data related to the monitored equipment. The calculated processed data can be used directly for evaluation or further calculated and analyzed to determine the condition of the monitored equipment.

[0063] The condition monitoring system according to the present invention may be a separate device directly connected to the monitored equipment, may be a device incorporated in the monitored equipment, or may be a separate device indirectly connected to the monitored equipment via a network such as a field network.

[0064] The measurement data used in the condition monitoring system according to the present invention is obtained by a data acquisition unit, such as a sensor attached to the monitored device or a measurement device that externally measures the condition of the monitored device. For example, the data acquisition unit may receive results received from the sensor in the form of sensor signals through a measurement unit included in the condition monitoring system and convert them into measurement data. Alternatively, the condition monitoring system may receive and acquire data in a format that can be used as measurement data via a network.

[0065] The processed data calculated by the status monitoring system involved in the present invention can be converted into a form that can be recognized by humans and output to be used as material for the person in charge to judge the status. Machine judgment can be made according to specified conditions, and it can also be fed back to the above-mentioned monitored equipment as statistical values.

[0066] use Figure 1 A basic configuration example of the state monitoring system 10 according to the present invention will be described. The state monitoring system 10 includes at least a measurement data processing unit 11 that calculates processed data from measurement data.

[0067] Furthermore, the state monitoring system 10 includes a data acquisition unit 12 for acquiring the measurement data processed by the measurement data processing unit 11. Figure 1 In FIG. 1 , an example of a method of receiving a sensor signal and converting it into measurement data to obtain the measurement data is shown, but for example, a network interface may be used to receive the measurement data from the outside.

[0068] As the measurement data, for example, time-series data representing any one of the physical quantities of the monitored device, such as jerk, angular jerk, acceleration, angular acceleration, displacement, angular displacement, position, phase, or pressure, can be used. However, these are merely examples, and the measurement data that can be used in the present invention to monitor the status of the monitored device is not particularly limited, as long as it is a physical quantity that can change over time.

[0069] Combine Figure 2 The following describes a first example flow of the sequence for calculating processed data from measurement data in the condition monitoring system 10 according to the present invention. In this first example flow, the measurement data processing unit 11 receives and processes the measurement data to be processed. A temporary storage unit (not shown) with sufficient capacity is required to store the measurement data to be processed during a single operation. The temporary storage unit can be either non-volatile or volatile memory. In this first example flow, data acquisition and processing do not require real-time performance, so even if response time increases, this does not hinder the processing itself.

[0070] First (S101), the data acquisition unit 12 receives measurement data from the outside such as a sensor (S102). As long as the measurement data is time series data along the time series, it can be processed directly. Although not shown in the figure in the process, at the stage of obtaining the measurement data, when the measurement data is not in the form of a time series, a step of converting it into first processed data along the time series is executed, and then the converted first processed data is used as time series data for subsequent processing. The number of data in the time series data is preferably at least 1,000, and more preferably more than 10,000. The spectrum is obtained by the high-speed Fourier transform described later, so if the number of data is too small, it is difficult to fully ensure the accuracy of the judgment. An example of the time series data is as follows. Figure 3 The horizontal axis is the time axis, and the vertical axis is the voltage received as a signal.

[0071] A subdivision step (S103) is performed to subdivide the above-mentioned time series data in the time direction to obtain subdivided data. Here, the subdivision is preferably performed at least 10 times, more preferably at least 100 times, and even more preferably at least 1000 times. As a specific subdivision method, the subdivision can be performed so that the number of data in each subdivided data set reaches a predetermined number of data, or the subdivision can be performed so that the number of subdivided data reaches a predetermined number of divisions. In addition, the number of divisions set here is proportional to the maximum frequency that can be grasped during processing in the diagnostic unit described later. Therefore, the number of divisions must be ensured to such an extent that the maximum frequency includes the frequency that is characteristic of the monitored device. In addition, the number of data in the subdivided data set here is proportional to the frequency resolution when extracting a specific frequency band. Therefore, the number of data in the subdivided data must be ensured to such an extent that the target frequency band can be extracted.

[0072] Furthermore, during fine segmentation, each sub-segmented data set can be non-overlapping, or a portion of the data can overlap with adjacent sub-segmented data in the time series. Alternatively, data not included in the sub-segmented data set can exist between adjacent sub-segmented data sets in the time series. Overlapping preserves information about a portion of the spectrum lost during segmentation between sub-segmented data sets. The overlap ratio is preferably 1% or greater for each sub-segmented data set, and more preferably 20% or greater. If the overlap ratio is too low, the effect of overlap in preventing information loss can be insufficient. Furthermore, a 1% overlap means that the total commonality with the preceding and following sub-segmented data sets in the time series is 1%, and the commonality with the preceding and following sub-segmented data sets is 0.5% each. Alternatively, the overlap ratio can be 180% or less for each sub-segmented data set, and preferably 120% or less. If the overlap exceeds 180%, processing efficiency will be significantly reduced. Here, an overlap ratio of 100% means that each point in the time series data set is included twice in the sub-segmented data set. If it exceeds 100%, a portion of the data set is included three times in the sub-segmented data set.

[0073] Next, each subdivided data item is processed individually. The following two steps process each subdivided data item independently. Therefore, the steps (S112 and S113) for the subdivided data item in the selected time series interval (S111) can be processed sequentially or in parallel for each subdivided data item. Furthermore, the selection of subdivided data items need not necessarily be performed in time series. However, after processing all subdivided data items, they are ultimately reassembled in time series. Therefore, selecting each subdivided data item in time series is preferred because it simplifies the management of the loop process and is therefore more convenient.

[0074] The FFT step (S112) is performed on the finely segmented data of the selected time series interval (S111) to calculate the spectrum data using a fast Fourier transform. Conventional algorithms can be used for the fast Fourier transform. In the present invention, the number of data processed in a single fast Fourier transform is reduced, allowing for efficient fast Fourier transform even in compact, power-efficient devices.

[0075] In addition, as the spectrum data, any form of an amplitude spectrum, a power spectrum, or a power density spectrum can be appropriately selected according to the required situation.

[0076] The spectral data obtained by performing the FFT step is subjected to a statistical step (S113) of extracting a partial spectrum of a certain frequency band and calculating the statistics of the partial spectrum. Specifically, this statistical step is divided into two stages. In the first stage, the partial spectrum of a certain frequency band within the full frequency band of the spectral data is extracted.

[0077] Here, as part of the frequency band to be extracted, specifically, it is effective to extract a limited frequency band that contains frequencies with a high frequency of characteristic behavior detected when a fault or abnormality occurs as a physical quantity of the monitored device. As a useful range, any frequency band within the range of 10Hz to 10kHz can be listed. If it is further expanded, any frequency band within the range of 0.1Hz to 30kHz can be listed. Partially selecting frequencies before and after the frequencies that are easy to grasp characteristic behavior within this range. However, it is not limited to this, and frequency bands that are not known frequency bands can also be extracted so that unexpected situations can be detected when they occur. Therefore, it is preferable to be able to arbitrarily change and set the extracted frequency band according to the type of monitored device, the purpose of monitoring, etc. In addition, there can be multiple frequency bands to be extracted. An example of the process of extracting multiple frequency bands will be described later.

[0078] As a final step in the statistical process, statistics of the extracted partial spectrum are calculated. Here, the statistics can be appropriately selected from values ​​that constitute the partial spectrum, such as the maximum value, total value, average value, and root mean square (RMS). These values ​​are not particularly limited, as long as they can be calculated.

[0079] The above S112 and S113 are performed on all finely segmented data, and then the next integration step (S122) is performed.

[0080] In the integration step (S122), the statistics obtained for all the fine segmentation data are arranged and integrated according to the time series of the original fine segmentation data, and are used as the processed data output by the measurement data processing unit. However, the statistics integrated along the time series in the integration step (S122) use the same statistics. The maximum value is not used as the statistic for the first fine segmentation data and the average value is used as the statistic for the other fine segmentation data. If it is the maximum value, the maximum value is arranged and integrated. That is, when performing Figure 2 In the case of a process, the selected statistics are common from the processing of the first fine-division data to the processing of the fine-division data of the final interval.

[0081] However, it is also possible to calculate multiple statistics in the statistical step (S113), and integrate each statistic in the integration step to obtain processed data for each statistic. For example, it is also possible to calculate the maximum value and the average value in the statistical step (S113), and integrate the maximum value and the average value into different processed data in the integration step (S122). Regardless of which process is used, it is preferred that the capacity of the memory used is smaller. However, it is also possible to temporarily store large-capacity spectrum data by storing the data in a flash memory or PSRAM, and the capacity of the temporarily stored data will not be such a big problem. In addition, the phase information of the spectrum data is reduced compared to the original time series data, so the capacity is halved. On the other hand, in the process of re-performing the FFT process, the consumption of CPU resources and RAM is relatively large, so it is difficult to implement in an inexpensive device, and is not preferred in the present invention.

[0082] Figure 4 The image shows the changes in the spectrum data, partial spectrum, and statistics of each data after fine segmentation based on the original time series data. In fact, the fast Fourier transform, extraction, and statistics calculation are performed for each interval in the time series. Figure 4 This is a hypothetical diagram showing what kind of image will appear after each fine segmentation data on the time series is finally integrated and the spectrum data and statistics calculated therefrom are arranged in the time series at each stage. Figure 3The same time series data. The second row of figures is the spectrum data of each fine segmentation data. The bars arranged vertically in the figure are the spectrum data of each fine segmentation data after fine segmentation. The vertical axis represents the frequency, and the color concentration represents the value of the spectrum. The horizontal axis is the time series in which the multiple spectrum data are virtually arranged. The third row of figures is a figure that only extracts a part of the frequency band in the vertical axis, i.e., the frequency direction, of the spectrum data in the second row. That is, similar to the second row, each bar is a partial spectrum, the vertical axis represents the frequency, and the color concentration represents the value of the spectrum. The horizontal axis is the time series in which the extracted partial spectrum data are virtually arranged. The fourth row is a figure in which statistics are calculated for each partial spectrum in the third row and arranged in time series. That is, the fourth row is the processed data after the integration step.

[0083] The reduction rate of the data count in the above process is roughly as follows. First, from the first to the second row, the total data count is halved due to the lack of phase information. During the extraction phase preceding the statistical step from the second to the third row, the data count is reduced from a fraction to a few tenths. Furthermore, during the statistical step following the third to the fourth row, the data count is reduced from a fraction to a few tenths, as the partial spectrum is converted into a single piece of data, such as the maximum value and average value.

[0084] The frequency bands extracted in the previous stage of the statistical step are preferably arbitrarily set and changeable. For example, if characteristic behavior can be detected in multiple frequencies A and B in separate frequency bands, a frequency band containing frequency A and another frequency band containing frequency B can be extracted from the spectral data obtained from the same time series data. Statistics are calculated for each in the statistical step, and processed data for each is obtained in the integration step. This outputs processed data containing information for frequency A and processed data containing information for frequency B. By being able to change the extracted frequency bands, the condition monitoring system can use the processed data to obtain the judgment results described below, thereby monitoring multiple possible conditions in parallel.

[0085] Combine Figure 5 and Figure 6The second and third process examples of the condition monitoring system 10 according to the present invention, in which the order of calculating processed data for multiple frequency bands based on measurement data as described above, are described. Prior to the start stage (S101), the predetermined multiple frequency bands to be selected and their order are set, and the measurement data processing unit 11 can read this setting. The information for the set frequency bands is pre-recorded in a storage unit (not shown), and is preferably set by external input as a command or by receiving and reading data. The process of receiving measurement data (S102) and performing the FFT step is the same as the first process. Then, after FFT processing of each subdivided data or all subdivided data, the frequency band to be extracted is specified from the set frequency band information.

[0086] exist Figure 5 In the second process shown, each time finely segmented data is selected, multiple frequency bands are sequentially specified. First, similar to the first process, finely segmented data is selected (S111), and the FFT step (S112) is performed. Next, a frequency band is specified (S141), and the statistics step for that band is performed (S113). These steps are repeated until the final frequency band is reached (S143 → No → S141). After the statistics step for all frequency bands for that finely segmented data is completed (S143 → Yes), the same process is repeated for the next finely segmented data (S121 → No → S111). In this second process, the spectral data stored in memory is only the finely segmented data being processed, making it easy to implement even in devices with limited memory capacity.

[0087] exist Figure 6 In the third process shown, after all subdivided data has been FFT-processed, multiple frequency bands are sequentially specified. First, similar to the first process, subdivided data is selected (S111) and the FFT step is performed (S112). The statistical step is then postponed, and subdivided data is sequentially selected and FFT steps are performed until the final interval is reached (S121 → No → S111). After the FFT steps for all intervals are completed (S121 → Yes), the first of the multiple frequency bands is specified (S141), and the statistical step (S113) and integration step (S122) are performed. These steps are repeated until the final frequency band is reached (S143 → No → S141). This third process requires temporary storage of spectral data for the entire interval of received measurement data, requiring a corresponding amount of memory capacity. This third process is particularly useful when the processed data is large (i.e., when the number of segments is large) or when a large number of frequency bands and statistics need to be extracted. Since the processed data can be sequentially outputted according to the type, it is not necessary to store all the processed data, and thus consumption of memory capacity can be suppressed.

[0088] also, Figure 5 and Figure 6 The embodiment in which the frequency band is specified during the second and third processes is merely an example, and the first frequency band may be specified from the beginning. In addition, the frequency band to be extracted in subsequent processes may be sequentially changed.

[0089] In the second and third flow examples, the processing content of the statistical step itself is the same as that of the first flow. The integration step of arranging the statistical values ​​in time series is also the same as that of the first flow.

[0090] The descriptions of the first to third flow examples above illustrate an example in which the measurement data processing unit 11 processes the measurement data to be processed after receiving it all at once, that is, an example in which the subdivision step is initiated after all the measurement data has been acquired. However, this is not limiting; the condition monitoring system according to the present invention may also initiate the subdivision step before the measurement data processing unit 11 has acquired all the measurement data. In this case, the basic sequence is the same as in the first to third flow examples above, but the subdivision step in S103 is performed only on the measurement data received at the start time. While executing steps S111 to S113, the remaining measurement data is acquired in parallel. Each time a new subdivision of measurement data is acquired, it is divided into new subdivisions. If measurement data cannot be acquired in time, the system temporarily enters a waiting state during the selection of the next subdivision of measurement data in S111. After the subdivision of measurement data to be processed is generated, the next FFT step (S112) and statistical step (S113) are performed. To perform this operation sequentially, it is best to determine whether the final interval has been reached (S121) not based on whether subdivided data remains in the memory, but rather on whether subdividing has completed up to the final interval of the scheduled measurement data to be acquired and whether the final interval has been read. By sequentially processing the subdivided data in parallel with measurement, the area required to store time-series data can be reduced. However, if the process of calculating statistics from the subdivided data is slower than acquiring the time-series data, a correspondingly larger amount of time-series data must be stored.

[0091] The state monitoring system 10 of the present invention can output the processed data obtained by the measurement data processing unit 11 directly or after obtaining a judgment result, thereby monitoring the state of the monitored device through an observer who observes the output and a device that receives the output. Compared with the original measurement data, the number of data in the processed data is greatly reduced, making it easier to read. However, even so, the number of data items may become tens to thousands along the time series. In the state of such processed data, it may be difficult to judge information. Therefore, the state monitoring system 10 preferably has a diagnostic unit 13 that uses the processed data to calculate a judgment result as a statistical value or judgment value. The judgment result is a numerical value or signal that can determine a certain state of the above-mentioned monitored device.

[0092] The processing performed by the diagnostic unit 13 can be implemented in various ways depending on the type of monitored device and the desired judgment result. Preferably, a second FFT step is performed to calculate second spectrum data using a fast Fourier transform based on the processed data. In the second spectrum data, the number of data points has been reduced by the measurement data processing unit 11, making it easier to detect feature quantities for frequencies remaining in the processed data. Preferably, the diagnostic unit 13 performs a judgment step based on this second spectrum data, calculating a judgment result as a statistical value or judgment value.

[0093] use Figures 7 to 14 An example of such a determination procedure in the diagnosis unit 13 will be described. Figure 7 The example of the second spectrum data after the processed data is subjected to the fast Fourier transform is shown. In this spectrum, the frequencies f representing the upper four peaks are n (f1 to f4) are shown as statistical values. The case where the feature can be found in the frequency itself, the case where the feature can be found in the order of the frequencies, etc. can be listed as an embodiment. In addition, the following embodiment can be listed: any one f representing the frequency of such a peak n Is there a characteristic frequency f in which the spectrum value tends to increase in the observation of the monitoring target device? f Near, for example, 0.95×f f ~1.05f f A true / false judgment value within the range of is used as the judgment result.

[0094] In addition, if Figure 8 As shown, an embodiment can be exemplified in which the maximum value (max) and the root mean square (rms) of the second spectrum data are calculated and their statistical values ​​are used as the determination result.

[0095] And, as Figure 9As shown, an embodiment can be exemplified in which a limited range before and after the peak of the second spectrum data is extracted, a root mean square (RMS) is calculated using the extracted data, and the statistical value is used as the determination result.

[0096] Furthermore, if Figure 10A As shown in FIG, an embodiment of obtaining a spectrum in which the second spectrum data is weighted can be listed. The weighting is related to Figure 10B The multiplication of window functions before and after the specific frequency is emphasized as shown. The window function shown here is only an example, and an appropriate window function can be selected according to the situation. Here, an embodiment can be used to select frequencies with a strong trend in the physical quantity of the monitored target device as the weighted frequencies.

[0097] Furthermore, the diagnostic unit 13 can also diagnose the status of the monitored device based on the spectrum values ​​and statistics, and use the result (normal, caution, warning, abnormal, stopped, etc.) as the judgment result. Examples of diagnostic methods include comparison with predetermined thresholds, trend analysis, and machine learning.

[0098] Furthermore, the diagnostic unit 13 can also perform quantitative processing on the processed data to obtain a judgment result without performing the second FFT step. The processed data has a significantly reduced number of data points compared to the original time-series data. Therefore, even if bandpass filtering, squaring, or subtraction, which would be computationally expensive and difficult to perform on the original time-series data, can be performed at a level that can be installed even on power-saving terminals. Statistics can also be directly calculated as judgment results without the use of such a fast Fourier transform.

[0099] For example, Figure 11 The results of bandpass filtering of the processed data are shown in Figure 12 In addition, Figure 11 The results of square processing of the processed data shown are as follows Figure 13 And, after subtracting the noise level, the result is as follows Figure 14 As shown. Furthermore, the noise-subtracted level refers to the value obtained by subtracting the noise-equivalent average value from the entire data, or by subtracting the minimum value from the entire data. Alternatively, the processed data itself may be used as the judgment result. Alternatively, a statistical value such as the peak value or root mean square (RMS) of the difference obtained by subtracting the minimum value from the maximum value in the processed data may be calculated.

[0100] The condition monitoring system 10 according to the present invention preferably includes an output unit 14, regardless of whether or not it includes a diagnostic unit 13. This output unit 14 is capable of outputting processed data output by the measurement data processing unit 11 and determination results output by the diagnostic unit 13 to a location outside the device or to a different component of the same device. The output destination of the output unit 14 can be a server within a local network, within a field network, or on the Internet, or a specific terminal.

[0101] A configuration example in which the condition monitoring system 10 according to the present invention is installed together with industrial machinery as a monitoring target device will be described.

[0102] exist Figure 15 In the first configuration example shown, the industrial machine 22 as the monitored device is provided separately from the state monitoring system 10. The edge device 21 constituting the state monitoring system 10 includes the aforementioned measurement data processing unit 11, data acquisition unit 12, diagnosis unit 13, and output unit 14. The edge device 21 is provided as a part of the state monitoring system 10 relative to the industrial machine 22 as the monitored device. Figure 15 Although only one industrial machine 22 and one edge device 21 are described, there can be multiple industrial machines 22 and multiple edge devices 21. For example, embodiments can be appropriately used to efficiently utilize sensor input from edge devices 21 in an environment where multiple industrial machines 22 are adjacent to each other, or to reflect analysis results from one industrial machine 22 to other industrial machines 22.

[0103] In this first structural example, the output unit 14 sends the output judgment result back to the control device 17 of the industrial machine 22, which is a monitored device that has a sensor 16 and sends measurement data to the edge device 21. Here, there are actually multiple sets of industrial machines 22 and edge devices 21, and each edge device 21 sends the judgment result back to the industrial machine 22 that sent the measurement data. The industrial machine 22 that receives the judgment result receives it through the internal control device 17. The control device 17 controls the industrial machine 22 based on the received judgment result. Examples of such control include stopping the corresponding industrial machine 22 when a judgment result indicating a fault occurs, or executing maintenance functions for the industrial machine 22. The communication standard between the industrial machine 22 and the edge device 21 is not particularly limited, and can be a wired connection or a wireless connection.

[0104] In this first configuration example, the output unit 14 transmits the output judgment results to the server 24 that constitutes the condition monitoring system 10. The judgment results can not only be directly utilized by each industrial machine 22, but can also be stored in the server 24 for statistical analysis and utilization. The condition monitoring system 10 may include multiple edge devices 21, but for the sake of aggregation and statistics, it is preferable that information from all edge devices 21 be centralized in the server 24 that collects this information. However, the server 24 can be a single server, a group of multiple servers, or a cloud server. The server 24 preferably includes a data storage unit 31 that receives and stores processed data from the output unit 14 of the edge device 21. Examples of the data storage unit 31 include, but are not limited to, long-term storage media such as magnetic disks and solid-state drives (SSDs). Furthermore, the server 24 preferably includes a data distribution unit 32 that outputs the processed data or further converted and edited data to the display device 25. Data distribution unit 32 can be a wired or wireless network interface, or it can use a video standard such as HDMI (registered trademark) or DisplayPort for display. Furthermore, if data distribution unit 32 is a network interface, it is also preferable to share the network interface when receiving processed data from the output unit 14. Furthermore, server 24 includes a control unit (not shown) that performs the aforementioned control.

[0105] In this first configuration example, the status monitoring system 10 includes a display device 25 that receives and displays the judgment results and data utilizing these judgment results received from the data distribution unit 32 of the server 24. The display device 25 displays information transmitted from the data distribution unit 32 of the server 24 in a manner that allows the person conducting the observation to primarily visually interpret the information. The display device 25 includes a data receiving unit 35 that receives the video signals and data transmitted from the data distribution unit 32, and a data display unit 36 ​​that serves as the actual display screen. Specifically, the display device 25 can be a monitor itself or a computer terminal equipped with a monitor. In the case of a monitor itself, the connection to the server 24 can be a video-standard cable. In the case of a computer terminal, the connection to the server 24 can be a network cable, a wireless antenna, or the like. Examples of devices used for display in the data display unit 36 ​​include LED lamps, cathode ray tubes, liquid crystal monitors, and organic EL monitors.

[0106] Furthermore, the state monitoring system 10 of the first configuration example may include devices and functions other than those described above. For example, an embodiment may include a speaker that emits a warning sound, a guidance sound, etc. instead of or in addition to the display device 25 .

[0107] As a first example of such a configuration, for example, the following system can be used: The wind turbines of a wind turbine generator system are monitored as industrial machinery 22, and edge devices 21 are installed on each wind turbine that constitutes the wind turbine generator system, thereby monitoring multiple wind turbines as a status monitoring system 10. The control device 17 of each wind turbine controls the wind turbine based on the judgment results received from the edge device 21. For example, this control can include emergency stop when an abnormality is detected, or output limitation based on the severity of the abnormality. Meanwhile, the server 24 that receives the judgment results stores the results in the data storage unit 31 and distributes them to the display device 25. The display device 25 displays the judgment results distributed from the server 24, thereby providing information to the manager of the wind turbine generator system.

[0108] Next, in Figure 16 In the second configuration example shown, a condition monitoring system 10 is built into an industrial machine 23, serving as the monitored device. Condition monitoring system 10 acquires (measures), processes, and diagnoses signals from sensors 16 installed on industrial machine 23. The system outputs the resulting determination to a control device 17, which forms part of industrial machine 23. Control device 17 uses the determination results to control industrial machine 23.

[0109] In addition, it can also be listed in Figure 16 In the second configuration example shown, the judgment result is output from the output unit 14 and the second configuration application structure is connected to the server 24 and the display device 25 in the first configuration example. For example, the display device 25 can also be a part of the industrial machine 23. Figure 17 In this case, the output unit 14 outputs to a portion corresponding to the data receiving unit 35 of the display device 25 according to the standard for display.

[0110] Also, it can be listed in Figure 16 The second configuration example shown in FIG. 1 is a further application mode of the second configuration in which the server 24 and the display device 25 are added. Figure 18 The output unit 14 serves as a network interface and transmits data to a server 24 located outside the status monitoring system 10. The server 24 is responsible for storing the data and transmitting it to a display device 25. This configuration allows data collection while avoiding an increase in the size of the industrial machine 23 itself.

[0111] In addition, next, Figure 19In the third configuration example shown, the condition monitoring system 10 transmits and receives signals via a field network 30 (FN30). The condition monitoring system 10 receives sensor signals from sensors installed on industrial machinery 22 via the field network 30, acquires (measures), processes, and diagnoses the signals. The determination results are transmitted (uploaded) to the field network 30 and then transmitted directly or after appropriate processing to the control device 17 of the industrial machinery 22. The control device 17 of the industrial machinery 22 controls the industrial machinery 22 based on the determination results received from the field network 30.

[0112] In addition, Figure 19 In the third configuration example shown, the display device 25 shown in the first configuration example can also be added as a part of the state monitoring system 10, and the judgment result can be sent from the output unit 14 to the data receiving unit 35, and the judgment result can be displayed on the data display unit. In addition, an application method can also be listed in which the display device 25 is set on the field network 30 to display the judgment result. This method is as follows Figure 20 shown.

[0113] Also, it can be listed in Figure 19 or Figure 20 The third configuration example shown further adds a server 24 connected to the field network 30 and a display device 25 connected to the server 24 to the application mode.

[0114] The status monitoring system of the present invention processes finely segmented data, significantly reducing the required computing resources compared to existing methods that centrally calculate all time series data. This makes it possible to monitor the status of target equipment even from inexpensive, power-efficient terminals such as edge devices. Its installation method is not limited to the aforementioned configuration example; other devices that utilize the resulting judgment results can also be added as appropriate, via a server, on-site network, directly connected, or built-in. However, it is preferred that devices directly connected to or built into the status monitoring system be added within a cost-effective and power-efficient range.

[0115] Description of Reference Numerals

[0116] 10…condition monitoring system; 11…measurement data processing unit; 12…data acquisition unit; 13…diagnosis unit; 14…output unit; 16…sensor; 17…control device; 21…edge device; 22…industrial machinery; 23…industrial machinery; 24…server; 25…display device; 30…field network; 31…data storage unit; 32…data distribution unit; 35…data receiving unit; 36…data display unit.

Claims

1. A condition monitoring system comprising a measurement data processing unit for calculating processed data based on measurement data, wherein: The measurement data processing unit performs: a fine segmentation step of finely segmenting time series data along a time series as the measurement data or as first processed data calculated based on the measurement data in a time direction; An FFT step, in which spectrum data is calculated by fast Fourier transform for each subdivided data after the time series data is subdivided; a statistical step of extracting a partial spectrum of a frequency band from the spectrum data and calculating a statistic of the partial spectrum; as well as An integration step, in which the statistics are arranged according to the time series of the original finely segmented data as the processed data.

2. The condition monitoring system according to claim 1, wherein: As the measurement data, time-series data of any one physical quantity of the monitored device, including jerk, angular jerk, acceleration, angular acceleration, displacement, angular displacement, position, phase, and pressure, is used.

3. The condition monitoring system according to claim 1 or 2, wherein: The device further comprises a diagnosis unit, which performs: A second FFT step, in which second spectrum data is calculated based on the processed data by high-speed Fourier transform; and A judgment step, in which a judgment result as a statistical value or a judgment value is calculated based on the second spectrum data.

4. The condition monitoring system according to claim 3, wherein: The statistical value in the determination step of the diagnosis unit is any one of a maximum value, a total value, an average value, and a root mean square value of the second spectrum data.

5. The condition monitoring system according to any one of claims 1 to 4, wherein: In the statistical step, the partial frequency band for the extraction may be changed to calculate a plurality of processed data having different partial frequency bands.

6. The condition monitoring system according to any one of claims 1 to 5, wherein: In the fine division step, the number of data of the finely divided data is set to a preset number, or the number of fine divisions of the finely divided data is set to a predetermined number of divisions.

7. The condition monitoring system according to any one of claims 1 to 6, wherein: The subdividing step is started before the measurement data processing unit acquires all the measurement data, or the subdividing step is started after the measurement data processing unit acquires all the measurement data.

8. The condition monitoring system according to any one of claims 1 to 7, wherein: In the fine division step, each of the fine division data is divided so as to overlap with other fine division data adjacent to each other in time series by a predetermined ratio, or the time series data is divided so as to be separated by a predetermined ratio and part of the time series data is discarded.

9. The condition monitoring system according to any one of claims 1 to 8, wherein: The spectrum data is any one of an amplitude spectrum, a power spectrum, and a power density spectrum.

10. The condition monitoring system according to any one of claims 1 to 9, wherein: The statistic is any one of a maximum value, a total value, an average value, and a root mean square value of the partial spectrum.

Images