JP7546502B2 - Condition diagnosis system - Google Patents

Description

The present invention relates to a condition diagnosis system.

Patent document 1 shows a bearing diagnostic device. In this diagnostic device, information for each bearing to be monitored is set via a parameter setting device, and a diagnostic judgment value is determined using this information.

JP 2002-022617 A

When monitoring a large number of devices, there is a problem in that if diagnostic settings need to be configured for each individual device, the configuration burden on the configuration staff becomes significant.

The present invention aims to provide a status diagnosis system that can diagnose the status of multiple monitored objects and has a simple setup process.

A condition diagnosis system according to one embodiment of the present invention comprises:
A condition diagnosis system for diagnosing conditions of a plurality of monitoring targets, comprising:
A setting unit that sets a common parameter common to the plurality of monitoring targets;
a determination unit that determines a plurality of individual parameters to be used in diagnosing states of the plurality of monitoring targets, based on the common parameter and values of a plurality of sensors that measure physical quantities of the plurality of monitoring targets;
a diagnosis unit that diagnoses a state of each of the monitoring objects based on the individual parameters corresponding to the monitoring object and a measured value of the physical quantity of the monitoring object;
a display unit that displays a diagnosis result obtained by the diagnosis unit; and
Equipped with
With the diagnostic result displayed on the display unit, the diagnostic result can be changed by changing a common parameter in the setting unit .

The present invention has the advantage of being able to provide a status diagnosis system that can diagnose the status of multiple monitored objects and has a simple setup process.

1 is a block diagram showing a configuration of a condition diagnosis system according to an embodiment of the present invention. 1 is a graph illustrating a reference model. 11 is a flowchart showing an example of a procedure from creation of a reference model to diagnosis. FIG. 13 is an image diagram showing an example of a common parameter setting screen. FIG. 11 is an image diagram showing a first display example of a diagnosis result. FIG. 11 is an image diagram showing a second display example of the diagnosis result. 5 is a flowchart showing a relearning process executed by the total management device and the status diagnosis device. 13 is a graph illustrating a reference model of a modified example. FIG. 13 is an image diagram showing a display example of a diagnosis result in a modified example.

The following describes an embodiment of the present invention in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a condition diagnosis system 1 according to an embodiment of the present invention. The condition diagnosis system 1 of this embodiment is a system that can diagnose the conditions of multiple monitored objects 101 without the user having to set thresholds or the like for each of the multiple monitored objects 101 in the system 100 to be diagnosed. The multiple monitored objects 101 may differ in model, type, operating conditions, operating environment, or multiple of these. In this embodiment, the multiple monitored objects 101 are multiple rotating devices (e.g., reducers, motors, etc.), but may also include various other devices such as actuators and motion transmission mechanisms. In this embodiment, the multiple monitored objects 101 are incorporated into one system 100, but may also be incorporated into different devices or different systems.

A sensor 102 that detects diagnostic information is attached to each monitored object 101, and a signal from the sensor 102 is sent to the condition diagnostic system 1. The sensor 102 measures, for example, the displacement or acceleration of vibration.

The conditions to be diagnosed include "normal", "caution", "warning", and "failure", which are divided into multiple stages ranging from normal conditions to abnormal conditions bordering on breakdowns. The order of deviation from the standard is "normal", followed by "caution", "warning", and finally "failure".

The condition diagnosis system 1 of this embodiment includes a condition diagnosis device 10 that performs diagnostic processing for each monitored object 101, and a general management device 20 that performs operation input from a user and output processing to the user.

The overall management device 20 is a computer equipped with a CPU (Central Processing Unit), a RAM (Random Access Memory) into which the CPU expands data, a storage device that stores the control program executed by the CPU, an input device (touch panel, keyboard, etc.) 31 through which operation commands are input by the user, a display device 32 capable of displaying images, and a communication module that communicates with the condition diagnosis device 10. In the overall management device 20, multiple functional modules are realized by the CPU executing the control program.

The multiple functional modules in the overall management device 20 include a common parameter setting processing unit 21 that performs setting input processing for common parameters, a common parameter storage unit 22 in which common parameters are stored, an operation processing unit 23 in which a user issues diagnosis-related commands and performs setting operations via an input device 31, a display processing unit 24 that displays and outputs diagnosis results on a display device 32, and a communication processing unit 25 that transmits and receives information related to diagnosis to and from the condition diagnosis device 10 via a communication module. The common parameter setting processing unit 21 corresponds to an example of a setting unit according to the present invention.

The common parameter storage unit 22 stores one or more common parameters depending on the state to be diagnosed. A common parameter is a parameter that is commonly used in the diagnosis of multiple monitoring targets 101, and corresponds to a threshold value of the state to be diagnosed. In this embodiment, three types of common parameters d1, d2, and d3 are stored, which respectively represent a threshold value between "normal" and "caution", a threshold value between "caution" and "warning", and a threshold value between "warning" and "failure".

The common parameter setting processing unit 21, for example, outputs an image having setting input fields for the three common parameters d1, d2, and d3 to the display device 32, and has the user input values into the setting input fields using the input device 31. Then, when the input is made, the common parameter setting processing unit 21 stores the input values in the common parameter storage unit 22.

The operation processing unit 23 accepts from the user via the input device 31 an operation to start a diagnostic process, an operation to start a learning process for determining individual parameters, an operation to output diagnostic results, and a selection operation to select which of multiple monitored objects 101 the diagnostic results of which should be displayed.

The condition diagnosis device 10 is a computer equipped with a CPU, a RAM in which the CPU expands data, a storage device in which a control program executed by the CPU is stored, an interface for inputting predetermined signals from the outside, and a communication module for communicating with the general management device 20. In the condition diagnosis device 10, multiple functional modules are realized by the CPU executing the control program.

The multiple functional modules in the condition diagnosis device 10 include a sensor signal acquisition unit 11 that captures the output (measurement value) of the sensor 102 via an interface, a communication processing unit 12 that performs communication processing with the overall management device 20 via a communication module, a memory processing unit 13 that has a common parameter storage unit 131, a reference model storage unit 132 including individual parameters, and a diagnosis result storage unit 133, and a calculation processing unit 14 that creates a reference model and diagnoses the condition of the monitored object 101 using the reference model. Values from the common parameter storage unit 22 of the overall management device 20 are copied to the storage units 131 of the multiple memory processing units 13.

The calculation processing unit 14 includes an AD conversion unit 141 that converts the signal of the sensor 102 acquired by the sensor signal acquisition unit 11 into a digital value, a filter processing unit 142 that extracts a waveform representing a feature from the digital value converted by the AD conversion unit 141, a feature calculation unit 143 that calculates the feature (effective value and peak value) from the extracted components, an MD calculation unit 144 that calculates the Mahalanobis' Distance (MD) of the feature, a reference model creation unit 145 that creates a reference model including individual parameters from a large number of feature values acquired during the learning period, and a diagnosis unit 146 that performs diagnosis processing on the feature values of the sample acquired during the diagnosis period. The reference model creation unit 145 corresponds to an example of a determination unit according to the present invention.

A plurality of sensor signal acquisition units 11 and storage processing units 13 are provided corresponding to the plurality of monitoring targets 101. The calculation processing unit 14 executes calculations for the plurality of monitoring targets 101 in parallel.

Next, the process for diagnosing one monitored object 101 will be described. In the condition diagnosis system 1 of this embodiment, the following process is executed in parallel for each of the multiple monitored objects 101.

<Samples and features>
The calculation processing unit 14 treats the output of the sensor 102 for a predetermined time length as one sample. Then, two feature quantities ( _x1 , _x2 ) are calculated from one sample by the processing of the AD conversion unit 141, the filter processing unit 142, and the feature quantity calculation unit 143. In this embodiment, the first feature quantity _x1 is the RMS (Root Mean Square: effective value) of the vibration displacement, and the second feature quantity _x2 is the peak value. The filter processing unit 142 is a calculation module that extracts a waveform that does not lose the feature quantities ( _x1 , _x2 ), and may be a low-pass filter or a band-pass filter, or may be a filter that extracts the envelope of the vibration waveform.

<Reference model, common parameters and individual parameters>
Fig. 2 shows a diagram for explaining the reference model. The reference model creation unit 145 creates a reference model MS1 using the feature quantities ( _x1 , _x2 ) of a large number of samples acquired during a learning period as original data. Hereinafter, the large number of samples acquired during the learning period will be referred to as a "reference sample group." In Fig. 2, a large number of plots P0 indicate the feature quantities ( _x1 , _x2 ) of the reference sample group.

The reference model MS1 is a space defined by the feature quantities ( _x1 , _x2 ) of the reference sample group acquired during the learning period, and has a dimension equal to the number of types of feature quantities (two, _x1 and _x2, in this embodiment). Furthermore, the reference model MS1 is a metric space model capable of calculating the Mahalanobis distance of any feature quantity ( _x1 , _x2 ). The Mahalanobis distance is defined by the average { _h1 , _h2 } of each feature quantity { _x1 , _x2 } in the reference sample group and the variance-covariance matrix Σ of the feature quantities in the reference sample group, and indicates the distance from the average point ( _h1 , _h2 ). Furthermore, the Mahalanobis distance is adjusted for each direction of the reference model MS1 according to the variance-covariance matrix Σ. That is, the distance is adjusted for each direction so that in a direction where the variance of the feature amount ( _x1 , _x2 ) of the reference sample group is large, the distance is relatively short even if the difference from the average point ( _h1 , _h2 ) is _large , and in a direction where the variance of the feature amount ( _x1 , _x2 ) of the reference sample group is small, the distance is relatively long even if the difference from the average point ( _h1 , h2) is small. By using such Mahalanobis distance, it is possible to express how much variation an arbitrary feature amount ( _x1 , _x2 ) deviates from the standard variation by assuming the variation of the feature amount ( _x1 , _x2 ) shown in the reference sample group as the standard variation (i.e., the degree of deviation from the standard).

The above average values {h ₁ , h ₂ } and variance-covariance matrix Σ correspond to one of the individual parameters included in the reference model MS 1 , and are used when determining the state of the monitored object 101 .

As shown in Fig. 2, the reference model MS1 may be accompanied by another type of individual parameter (boundaries e1, e2, e3) that functions as a threshold corresponding to the common parameter. The common parameter is a value set by a user or the like, and is a threshold distance (d1, d2, d3) for identifying a plurality of states ("normal", "caution", "warning", "failure") of the monitored object 101, and is expressed by the Mahalanobis distance. That is, if the Mahalanobis distance of the feature quantity ( _x1 , _x2 ) obtained during the diagnosis period is shorter than the threshold distance d1, it indicates a "normal" state, if it is equal to or greater than the threshold distance d1 and less than the threshold distance d2, it indicates a "caution" state, if it is equal to or greater than the threshold distance d2 and less than the threshold distance d3, it indicates a "warning" state, and if it is equal to or greater than the threshold distance d3, it indicates a "failure" state.

As shown in FIG. 2, the individual parameters corresponding to the common parameters correspond to the boundaries e1, e2, and e3 when the common parameters (threshold distances d1, d2, and d3) are applied to the reference model. In the reference model MS1 in FIG. 2, the distances from the average point (h ₁ , h ₂ ) to each point of the boundary e1 are the same distance (threshold distance d1). Similarly, the distances from the average point (h ₁ , h ₂ ) to each point of the boundary e2 are the same distance (threshold distance d2), and the distances from the average point (h ₁ , h ₂ ) to each point of the boundary e3 are the same distance (threshold distance d3). Therefore, the inside of the boundary e1 corresponds to a region where the Mahalanobis distance is shorter than the threshold distance d1, and the outside of the boundary e1 corresponds to a region where the Mahalanobis distance is longer than the threshold distance d1. The relationship between the threshold distances d2 and d3 and the boundaries e2 and e3 is also similar.

The common parameter is expressed by the Mahalanobis distance and indicates the deviation from the standard. Therefore, the common parameter can be a common value for a plurality of monitoring targets 101 having different models, models, operating environments, or operating conditions. On the other hand, the standard variation of the average score (h 1 , h 2 ) and the feature amount (x _{1 ,} x ₂ ) of the reference sample group acquired during the learning period and the reference model MS1 created from the feature amount (x ₁ , x ₂ ) of the reference sample group differs depending on the model, model, operating environment, or operating conditions of the monitoring target 101. While the reference model _MS1 differs depending on the monitoring target 101 in this way, the deviation of the reference model MS1 from the standard is indicated by the common parameter, and the individual parameters (boundaries e1, e2, e3) obtained by applying the common parameter to the reference model MS1 become values indicating the deviation of the reference model MS1 from the standard, which changes depending on the model, model, operating environment, or operating conditions of the monitoring target 101. Therefore, by using the individual parameters (boundaries e1, e2, e3), it is possible to identify the "normal", "caution", "warning", and "failure" states of individual monitored objects 101 having different models, types, operating environments, or operating conditions, etc.

The reference model creation unit 145 calculates the above individual parameters (mean score ( _h1 , _h2 ) and variance-covariance matrix Σ) for each of the multiple monitoring targets 101 from the feature amounts ( _x1 , _x2 ) of the reference sample group of each monitoring target 101, and creates multiple reference models MS1 corresponding to the multiple monitoring targets 101 respectively. The created reference model MS1 (mean score ( _h1 , _h2 ) and variance-covariance matrix Σ) is stored in the corresponding storage unit 132 of the memory processing unit 13. The reference model creation unit 145 may also calculate boundaries e1, e2, e3 by applying the common parameters in storage unit 131 to the reference model MS1, and store this in storage unit 132 as another individual parameter of the reference model MS1.

<Diagnosis process using reference model>
The diagnosis unit 146 diagnoses whether the monitored object 101 is in a "normal", "caution", "warning", or "failure" state based on the feature amounts ( _x1 , _x2 ) of each sample acquired during the diagnosis period, the reference model stored in the storage unit 132, and the common parameters stored in the storage unit 131. In the diagnosis process, the feature amounts ( _x1 , _x2 ) of the sample are determined based on where they are located in the reference model in relation to the boundaries e1, e2, and e3. If they are within the boundary e1, they are determined as "normal", if they are outside the boundary e1 and within the boundary e2, they are determined as "caution", if they are outside the boundary e2 and within the boundary e3, they are determined as "warning", and if they are outside the boundary e3, they are determined as "failure".

In practice, the MD calculation unit 144 calculates the Mahalanobis distance on the reference model for each sample feature ( _x1 , _x2 ) acquired during the diagnosis period. Furthermore, the diagnosis unit 146 compares the calculated Mahalanobis distance with the common parameters d1, d2, and d3 to determine where the sample feature ( _x1 , _x2 ) is located in the section partitioned by the boundaries e1 to e3, and determines the state of the monitored object 101.

In the above description, an example was given in which the common parameters are expressed by the Mahalanobis distance, and the reference model is a distance space model in which the Mahalanobis distance can be defined. However, the distance applied to the common parameters and the distance defined in the reference model are not limited to the Mahalanobis distance, and may be replaced by a distance having a similar characteristic. That is, if the distance is defined so that the difference from the average value ( _h1 , _h2 ) is large in the direction in which the variance of the feature amount ( _x1 , _x2 ) of the reference sample group is large but is not a relatively long distance value, and the difference from the average value ( _h1 , _h2 ) is small in the direction in which the variance of the feature amount ( _x1 , _x2 ) of the reference sample group is small but is a relatively long distance value, it can be applied in the same way as the Mahalanobis distance. That is, if the distance is defined so that the deviation from the standard is shown using the value and variation of the feature amount of the reference sample group as the standard, it can be applied in the same way as the Mahalanobis distance. Add f1, f2, f3, g1, g2, g3

In the condition diagnosis device 10, the acquisition of samples during the learning period and diagnosis period, the creation process of a reference model, and the diagnosis process using the reference model are performed in parallel for each of the multiple monitored objects 101.

<Processing flow>
Fig. 3 is a flowchart showing an example of a processing procedure from the creation of a reference model to diagnosis. Fig. 3 shows processing for one monitored object 101. In the condition diagnosis system 1, multiple sets of processing from the creation of a reference model in Fig. 3 to diagnosis are executed in parallel in correspondence with each of the multiple monitored objects 101.

When a learning period is set by a user or the like, the process of creating a reference model is started from step S1 during the learning period. The learning period refers to a period during which the system 100 to be diagnosed is operating normally, and is set to acquire original data (feature quantities ( _x1 , _x2 ) of the reference sample group) for creating a reference model for each monitored object.

When the reference model creation process is started, first, the AD conversion unit 141 converts the output of the sensor 102 into a digital value (step S1), and the filter processing unit 142 extracts a waveform that facilitates the calculation of the feature amount ( _x1 , _x2 ) (step S2). Furthermore, the feature amount calculation unit 143 samples the extracted waveform at predetermined time intervals, calculates the RMS of the vibration displacement for one sample as the feature amount _x1 , and calculates the peak value of the vibration displacement for one sample as the feature amount _x2 (step S4). In addition, when calculating the feature amount ( _x1 , _x2 ), the feature amount calculation unit 143 may perform an intermediate process such as frequency analysis of the waveform extracted in step S2 and selecting a waveform in a specific frequency band (step S3), and calculate one or both of the feature amounts _x1 and _x2 for the intermediately processed waveform. The intermediate process may be a process that can reduce waveforms that are not necessary for determining the state of the monitored object 101 and extract waveforms that have a large influence on the state determination. The feature amount calculation unit 143 accumulates the calculated feature amounts (x ₁ , x ₂ ) in the storage unit (step S5).

Then, the calculation processing unit 14 judges whether the learning period has ended (step S6), and repeats the processes of steps S1 to S5 until the learning period ends. Then, when the learning period ends, the reference model creation unit 145 calculates a reference model from the feature quantities (x ₁ , x ₂ ) of the accumulated many samples (reference sample group) (step S7). As described above, the reference model is a metric space model having the average point (h ₁ , h ₂ ) of the feature quantities (x ₁ , x ₂ ) of the reference sample group and the variance-covariance matrix Σ of the feature quantities (x ₁ , x ₂ ) of the reference sample group. The average value (h ₁ , h ₂ ) and the variance-covariance matrix Σ are individual parameters used to diagnose the state of the corresponding monitoring target 101. In addition, the reference model created in step S7 may include boundaries e1 to e3 (FIG. 2) obtained by applying the common parameters to the reference model as individual parameters. The reference model (individual parameters defining it) is stored in the storage unit 132 of the memory processing unit 13.

When the diagnostic process is started by a user or the like during the diagnosis period, the diagnostic process starts from step S11. The diagnostic period may be set to a period during which diagnosis is required, such as a period during which the system 100 to be diagnosed is in operation. When the diagnostic process starts, steps S11 to S14 are executed in the same manner as steps S1 to S4 described above, samples are acquired sequentially during the diagnosis period, and the feature quantities (x ₁ , x ₂ ) of each sample are calculated.

Then, when the feature quantity ( _x1 , _x2 ) of one sample is calculated, the MD calculation unit 144 uses the feature quantity ( _x1 , _x2 ) and the individual parameters of the reference model to calculate the Mahalanobis distance (MD) of the feature quantity ( _x1 , _x2 ), that is, the deviation degree from the standard in the reference model (step S15). Furthermore, the diagnosis unit 146 compares the deviation degree (Mahalanobis distance) with a common parameter (threshold value) to diagnose the state (step S16). Then, the diagnosis unit 146 stores the diagnosis result in the storage unit 133 and notifies the diagnosis result to the outside according to a predetermined condition (step S17). The diagnosis result is stored in the storage unit 133 of the memory processing unit 13. The stored diagnosis result includes, for example, the date and time, the monitoring target, the feature quantity ( _x1 , _x2 ), and each information of the determined state that are linked to each other.

The overall management device 20 determines whether or not a display request for the diagnosis result for any of the monitored objects 101 has been made (step S18). If such a request has been made, the display processing unit 24 reads out the diagnosis result for the monitored object 101 from the storage unit 133 of the memory processing unit 13 via the communication processing unit 25, converts it into a predetermined format (such as a graph display), and outputs it from the display device 32 (step S19). Furthermore, the overall management device 20 determines whether or not a command to end the diagnosis has been made (step S20). If such a command has not been made, the overall management device 20 causes the condition diagnosis device 10 to continue the diagnosis process, and the process is repeated from step S11. On the other hand, if a command to end the diagnosis has been made, a command to end the diagnosis is sent from the overall management device 20 to the condition diagnosis device 10, and the diagnosis process ends.

<Common parameter setting process>
4 shows an example of a common parameter setting screen. The setting screen H1 is created by the common parameter setting processing unit 21 and output to the display device 32. The user can display the setting screen H1 and set the common parameters by operating a common parameter setting request via the operation processing unit 23 of the overall management device 20. The setting screen H1 has input fields L1 to L3 into which the common parameters can be input, and the user can set the common parameters by inputting numerical values into the input fields L1 to L3 using the input device 31.

On the setting screen, when the learning period has elapsed, a graph G1 of the reference model may be displayed with one of the feature quantities ( _x1 , _x2 ) on the vertical axis and the other on the horizontal axis, and boundaries e1 to e3 corresponding to the common parameters may be indicated. By changing the value of the common parameter, the positions of the boundaries e1 to e3 may be changed in conjunction with each other. With such a display, the user can set the common parameters while considering how a change in the value of the common parameter will affect the diagnosis result of the state of the monitored object 101.

<Display of diagnosis results>
5 is a diagram showing a first display example of the diagnosis result. A user can specify one of the multiple monitoring targets 101 via the operation processing unit 23 of the total management device 20 and input a request to display the diagnosis result, thereby displaying a diagnosis result screen H2 of the monitoring target 101 on the display device 32.

The diagnostic result screen H2 is created by the display processing unit 24 and output to the display device 32. The first display example in FIG. 5 includes a reference model graph J1 showing the diagnostic result on a reference model, comparison graphs J2 and J3 comparing the changes in the feature amounts at the time of learning with those at the present time, and time series graphs J4 and J5 showing the diagnostic result in a time series.

The reference model graph J1 is a graph in which one of the features ( _x1 , _x2 ) is taken on the vertical axis and the other on the horizontal axis, and includes a plot P0 showing the features of the reference specimen group acquired during the learning period, individual parameters (boundaries e1, e2, e3) corresponding to the common parameters, and a measurement plot P1 showing the features of multiple specimens acquired during the diagnosis period. The user can grasp the state of the monitored object 101 from the relative positional relationship between the measurement plot P1 and the boundaries e1, e2, e3.

The comparison graphs J2 and J3 include a comparison graph J2 for one feature amount _x1 and a comparison graph J3 for the other feature amount _x2 . Each comparison graph J2 and J3 shows the transition of the feature amounts of a group of reference samples acquired during a learning period and the transition of the feature amounts of a plurality of samples acquired in the vicinity of the present. Based on the comparison graphs J2 and J3, the user can compare the transition of the reference feature amount with the transition of the current feature amount and grasp the difference between the reference time and the present of the monitored object 101.

The time series graphs J4 and J5 include a time series graph J4 for one feature quantity _x1 and a time series graph J5 for the other feature quantity _x2 . Each of the time series graphs J4 and J5 shows the transition of the feature quantity from a predetermined period before to the present, and the diagnosis result of the state at each time (deviation from the standard). The diagnosis result of the state is displayed so as to be identifiable, for example, by the type (color) of the plot. From the transition of the diagnosis result, the time series graphs J4 and J5 show time series information of the deviation from the standard. The user can grasp the transition of the state of the monitored object 101 over a certain period from the time series graphs J4 and J5.

In the first display example, the graphs J1 to J5 are output simultaneously, but one or more of the graphs J1 to J5 may be output, or each of the graphs J1 to J5 may be output individually. Also, by selecting one of the simultaneously output graphs J1 to J5, the selected graph may be enlarged and displayed.

FIG. 6 is a diagram showing a second display example of the diagnosis result. The diagnosis result screen H3 of the second display example is created by the display processing unit 24 and output to the display device 32. The second display example includes a time series graph J6 showing the deviation of the feature amount from the standard during the diagnosis period in a time series. The time series graph J6 is a graph showing the deviation (Mahalanobis distance: MD) of the feature amounts (x ₁ , x ₂ ) of a plurality of samples acquired from a predetermined period to the present in a time series. The time series graph J6 shows threshold lines k1 to k3 indicating the common parameters d1, d2, and d3. In addition, plots showing the deviation of the feature amounts (x ₁ , x ₂ ) at each time point are displayed with different types (colors) of each plot so that the states of the monitoring target 101, "normal", "caution", "warning", and "failure", can be identified. The user can grasp the transition of the state of the monitoring target 101 during a certain period from the time series graph J6.

<Displaying diagnostic results and modifying common parameters>
The overall management device 20 is configured to be able to execute a common parameter setting process in parallel with the output of the first or second display example of the diagnosis result. When the user changes the value of the common parameter while the diagnosis result is being displayed, the reference model creation unit 145 of the condition diagnosis device 10 recalculates the reference model in conjunction with the change, and further, the diagnosis unit 146 redoes the diagnosis again by associating the specimens acquired during the past diagnosis period with the new common parameters. Then, the data in each of the storage units 131 to 133 of the memory processing unit 13 of the condition diagnosis device 10 is rewritten.

When the data in each of the storage units 131 to 133 of the memory processing unit 13 is rewritten, the display processing unit 24 updates the display of the diagnosis results using the new reference model and diagnosis results. When the user changes the settings of the common parameters while the diagnosis results are being displayed, the boundaries e1 to e3 shown in the diagnosis results and the plots showing the diagnosis results in the time series graphs J4, J5, and J6 change in accordance with the change in the common parameters.

<Notification Processing>
The overall management device 20 may include an alarm processing unit 26 ( FIG. 1 ) that monitors the diagnosis results for all the monitored objects 101 stored in the storage unit 133 of the multiple memory processing units 13 of the status diagnosis device 10. When the diagnosis results for all the monitored objects 101 reach a preset state (e.g., a state in which two or more monitored objects 101 are determined to be "warning" or a state in which at least one monitored object 101 is determined to be "failure"), the alarm processing unit 26 notifies the user of the state by display output, audio output, or the like. The alarm processing unit 26 allows the user to easily grasp the state change of the system including the multiple monitored objects 101.

<Relearning process>
7 is a flowchart showing the re-learning process executed by the total management device 20 and the condition diagnosis device 10. The operation processing unit 23 of the total management device 20 repeatedly performs a process of determining whether or not there is a re-learning request from the user via the input device 31 (step S31), and if there is a re-learning request, specifies the monitored object 101 to be re-learned, the learning start timing, and the learning period, and commands the calculation processing unit 14 of the condition diagnosis device 10 to recreate the reference model (step S32). The learning start timing and the learning period may be specified by the user, or may be a timing and period that are predetermined from the time of the re-learning request.

When a re-learning command is issued, the calculation processing unit 14 acquires a standard sample group from the waveform acquired by the sensor signal acquisition unit 11 during the specified learning period, and calculates the feature quantities (step S33). Then, the reference model creation unit 145 creates a reference model using the feature quantities of the standard sample group, and stores the reference model in the storage unit 132 of the memory processing unit 13 (step S34). The method of creating the reference model is the same as the method described above.

The operation processing unit 23 of the comprehensive management device 20 may be capable of inputting an operation to request that the reference models of all monitored objects 101 be re-learned collectively, or may be capable of issuing a request to specify the reference model of each monitored object 101 and re-learn it.

By using the re-learning process, when one or more devices that are the monitored objects 101 in the system 100 to be diagnosed are replaced, or when the operating environment of the system 100 changes, the reference model can be updated in accordance with such changes in the situation, and subsequent diagnosis of the monitored object 101 can be continued based on the reference model that is appropriate to the situation.

The re-learning process is not limited to being executed based on user operations, but may be configured to be executed based on various pre-set conditions.

As described above, according to the condition diagnosis system 1 of the present embodiment, the reference model creation unit 145 creates a reference model from the value of the sensor 102 (specifically, a group of reference samples acquired during the learning period) that measures a predetermined physical quantity (vibration displacement) of the monitored object 101. Furthermore, during the diagnosis period, the diagnosis unit 146 diagnoses the state of each monitored object 101 using individual parameters (the average value (h ₁ , h ₂ ) and the variance-covariance matrix Σ of the feature quantities) that define the reference model, the common parameters d1 to d3, and the value of the sensor 102 (any sample acquired during the diagnosis period). Therefore, even if the system 100 to be diagnosed includes multiple monitored objects 101 with different models, models, operating environments, and operating conditions, the states of the multiple monitored objects 101 can be diagnosed by setting the common parameters d1 to d3 common to the multiple monitored objects 101. Therefore, the load of the setting process required to diagnose the multiple monitored objects 101 can be significantly reduced.

Furthermore, according to the condition diagnosis system 1 of this embodiment, a display processing unit 24 is provided that displays the diagnosis results of the diagnosis unit 146 on the screen of the display device 32. Therefore, the user can understand the condition of the monitored object 101 from the display of the diagnosis results.

Furthermore, according to the condition diagnosis system 1 of this embodiment, the display of the diagnosis result includes information indicating the degree of deviation of the state of the monitored object 101 from the standard. Therefore, the user can understand from the display of the diagnosis result how much the state of the monitored object 101 has deviated from the standard.

Furthermore, according to the condition diagnosis system 1 of this embodiment, the display of the diagnosis result (time series graphs J4, J5) includes time series information on the deviation degree. Therefore, the user can understand from the display of the diagnosis result how much the state of the monitored object 101 has deviated from the standard and the progress along a time series.

Furthermore, according to the condition diagnosis system 1 of this embodiment, the display of the diagnosis result (reference model graph J1) shows the boundaries e1 to e3 obtained by applying the common parameters to the reference model. Therefore, the user can understand from the display of the diagnosis result what threshold the generalized common parameters correspond to in the diagnosis of each monitored object 101.

Furthermore, according to the condition diagnosis system 1 of this embodiment, when the common parameters are changed, the positions of the boundaries e1 to e3 in the display of the diagnosis result (reference model graph J1) change in accordance with the change in the common parameters. Therefore, the user can change the common parameter settings while taking into consideration how the common parameters affect the diagnosis result.

Furthermore, according to the condition diagnosis system 1 of this embodiment, the sensor signal acquisition unit 11 acquires a signal from the sensor 102 that detects vibrations of the monitored object 101, and the condition of the monitored object 101 is diagnosed based on this signal. Therefore, for example, when the monitored object 101 is a device that includes moving parts and whose vibrations increase as it deteriorates, it is possible to accurately diagnose the condition of the monitored object 101.

Furthermore, according to the condition diagnosis system 1 of this embodiment, the reference model can be recreated by a re-learning process. In other words, the individual parameters used to diagnose each monitored object 101 can be redetermined. Therefore, for example, when a part of the monitored object 101 is replaced or the operating environment changes significantly, the reference model can be recreated to suit the new situation, and diagnosis processing adapted to the new situation can be continued.

Furthermore, according to the condition diagnosis system 1 of this embodiment, the reference model creation unit 145 creates a reference model using statistics calculated from the signal value of the sensor 102, while the statistics include only the feature x ₁ (RMS of the vibration displacement of the monitored object 101) and the feature x ₂ (peak value of the vibration displacement). The RMS represents the average amount of the magnitude of the vibration displacement. By using only such statistics (feature x ₁ , x ₂ ), the reference model can be expressed in two dimensions, and by narrowing down the diagnostic materials to two feature amounts, the user can easily grasp the principle of diagnosis. Furthermore, by the diagnostic process using the above two feature amounts, accurate diagnostic results can be obtained for multiple monitored objects 101 with different types, models, operating environments, and operating conditions.

Furthermore, according to the condition diagnosis system 1 of this embodiment, the display of the diagnosis result includes a reference model graph J1 in which one of the feature quantities _x1 and _x2 is plotted on the vertical axis and the other on the horizontal axis. The reference model graph J1 is a graph that shows the diagnosis result of the state of the monitoring target 101 from the relative positional relationship between the feature quantities ( _x1 , _x2 ) obtained during the diagnosis period and the boundaries e1 to e3 indicating threshold values. Therefore, the user can understand from the reference model graph J1 what state the monitoring target 101 is in and how it is being diagnosed.

(Modification)
8 is a graph illustrating a reference model of the modified example. The modified example is similar to the above-described embodiment except that the individual parameters created by the reference model creation unit 145 and the diagnosis method of the diagnosis unit 146 are different.

First, the reference model creation unit 145 uses a group of reference samples acquired during the learning period to create a metric space model with a defined Mahalanobis distance as the basis for the reference model MS2, similar to the reference model MS1 described above.

Furthermore, the reference model creation unit 145 obtains boundaries e1, e2, e3 obtained by applying the common parameters d1, d2, d3 to the reference model, and calculates the maximum value in the axial direction of the feature amount _x1 of each boundary e1, e2, e3 as first threshold values f1, f2, f3, and the maximum value in the axial direction of the feature amount _x2 of each boundary e1, e2, e3 as second threshold values g1, g2, g3. In the modified example, the first threshold values f1, f2, f3 and the second threshold values g1, g2, g3 correspond to individual parameters used when diagnosing the state ("normal", "caution", "warning", "failure") of the monitored object 101.

The diagnosis unit 146 compares one feature amount _x1 of the sample acquired during the diagnosis period with first thresholds f1 to f3, and compares the other feature amount _x2 with second thresholds g1 to g3, and diagnoses the state of the feature amount ( _x1 , _x2 ) that has a higher deviation from the standard as the state of the monitoring target 101. That is, the diagnosis unit 146 determines a provisional "normal" state if the feature amount x1 is less than the first threshold value f1, a provisional "caution" state if the feature amount _x1 is equal to or greater than the first threshold value f1 and less than the first threshold value f2, a provisional "warning" state if the feature amount x1 is equal to or greater than the first threshold value f2 and less than the first threshold value f3, and a provisional "failure" state if the feature amount x1 is equal to or greater than the first threshold value f3. Furthermore, the diagnosis unit 146 determines a provisional "normal" state if the feature amount _x2 is less than the second threshold value g1, determines a provisional "caution" state if the feature amount x2 is equal to or greater than the second threshold value g1 and less than the second threshold value g2, determines a provisional "warning" state if the feature amount x2 is equal to or greater than the second threshold value g2 and less than the second threshold value g3, and determines a provisional "failure" state if the feature amount x2 is equal to or greater than the second threshold value g3. Then, the diagnosis unit 146 diagnoses the worse of the state diagnosed for the feature amount _x1 and the state diagnosed for the feature amount _x2 as the state of the monitoring target 101.

When one feature amount _x1 exceeds the first threshold value f1, or when the other feature amount _x2 exceeds the second threshold value g1, the points on the reference model MS2 corresponding to the two feature amounts ( _x1 , _x2 ) are located in an area outside the boundary e1. Therefore, when the monitored object 101 is a device in which the feature amounts _x1 , _x2 are expected to change in a direction to increase as the state of the monitored object 101 transitions from "normal" to "failure", the state of the monitored object 101 can be correctly diagnosed by the diagnosis method using the first threshold values f1 to f3 and the second threshold values g1 to g3, as in the above-mentioned embodiment.

Figure 9 is an image diagram showing an example of the display of the diagnosis result in the modified example. The display of the diagnosis result is created by the display processing unit 24 and output to the display device 32. The display example of the modified example includes a reference model graph J11 showing the diagnosis result on a reference model, and time series graphs J12 and J13 showing the diagnosis deterioration in a time series.

The reference model graph J11 is a graph in which one of the feature quantities ( _x1 , _x2 ) is taken on the vertical axis and the other on the horizontal axis, and shows a plot P0 showing the feature quantities of the reference specimen group acquired during the learning period, first threshold values f1-f3 and second threshold values g1-g3 (individual parameters) calculated from boundaries e1, e2, e3 corresponding to the common parameters, and a measurement plot P1 showing the feature quantities of a plurality of specimens acquired during the diagnosis period. The user can grasp the state of the monitoring target 101 from the relative positional relationship between the measurement plots and the first threshold values f1-f3 and second threshold values g1-g3.

The time series graphs J12 and J13 include a time series graph J12 for one feature amount _x1 and a time series graph J13 for the other feature amount _x2 . The time series graph J12 shows a plot showing the transition of the feature amount _x1 from a predetermined period ago to the present, and first thresholds f1 to f3. The time series graph J13 shows a plot showing the transition of the feature amount _x2 from a predetermined period ago to the present, and second thresholds g1 to g3. The positional relationship between the first thresholds f1 to f3 and the plot of the feature amount _x1 represents the deviation of the feature amount _x1 from the standard, so the time series graph J12 shows time series information on the deviation of the feature amount _x1 from the standard. Similarly, the positional relationship between the second thresholds g1 to g3 and the plot of the feature amount _x2 represents the deviation of the feature amount _x2 from the standard, and therefore the time series graph J13 shows time series information of the deviation of the feature amount _x2 from the standard. The user can see the transition of the relative positional relationship between the first thresholds f1 to f3 or the second thresholds g1 to g3 and each plot from the time series graphs J12 and J13, and grasp the transition of the state of the monitored object 101 along the time series.

<Displaying diagnostic results and modifying common parameters>
The total management device 20 is configured to be able to execute a common parameter setting process in parallel with the display of the diagnosis results. When the user changes the value of the common parameter while the diagnosis results are being displayed, the reference model creation unit 145 of the condition diagnosis device 10 recalculates the first threshold values f1-f3 and the second threshold values g1-g3 of the reference model in conjunction with the change, and the diagnosis unit 146 redoes the diagnosis of the feature quantities (x ₁ , x ₂ ) at each time point during the diagnosis period previously calculated. Then, the reference model creation unit 145 and the diagnosis unit 146 rewrite the data in the storage units 132, 133 of the memory processing unit 13.

When the data in the storage processing unit 13 of the condition diagnosis device 10 changes, the display processing unit 24 updates the display of the diagnosis result using new first thresholds f1 to f3 and second thresholds g1 to g3. Thus, when the user changes the settings of the common parameters while the diagnosis result is being displayed, the positions of the first thresholds f1 to f3 and second thresholds g1 to g3 (individual parameters) shown in the diagnosis result are changed to match the new common parameters.

As described above, according to the state diagnosis system 1 of the modified example, by setting the common parameters d1 to d3 common to the multiple monitored objects 101, the reference model creation unit 145 creates a reference model including individual parameters (first thresholds f1 to f3, second thresholds g1 to g3) used when individually diagnosing each monitored object 101 from the value of the sensor 102 (specifically, a reference sample group acquired during a learning period) that measures a predetermined physical quantity (vibration displacement) of the monitored object 101 and the common parameters. During the diagnosis period, the diagnosis unit 146 diagnoses the state of the monitored object 101 using the feature quantities x ₁ and x ₂ calculated from the value of the sensor 102 and the individual parameters (first thresholds f1 to f3, second thresholds g1 to g3) of the reference model. Therefore, even if the system 100 to be diagnosed includes multiple monitored objects 101 with different models, models, operating environments, and operating conditions, by setting the common parameters d1 to d3 common to the multiple monitored objects 101, diagnosis of the state of the multiple monitored objects 101 can be realized by setting the common parameters. Therefore, the load of the setting process required to diagnose a plurality of monitoring targets 101 can be significantly reduced.

In the state diagnosis system 1 of the modified example, the reference model creation unit 145 creates a reference model including individual parameters (first thresholds f1 to f3, second thresholds g1 to g3), but the calculation method is not limited to this. For example, the reference model creation unit 145 may obtain the average value or median value of each sample group (one feature amount x ₁ of the sample acquired in the diagnosis period and the other feature amount x ₂ ) from the reference sample group acquired in the learning period as individual parameters, and may multiply the average value or median value by a predetermined coefficient based on the individual parameters to set the first thresholds f1 to f3 and the second thresholds g1 to g3. The above-mentioned predetermined coefficient may be included in the common parameters, may be determined individually for each model, or may be changed later. The learning period may be a common value for multiple monitoring targets 101 having different models, models, operating environments, or operating conditions, and may be included in the common parameters. In the above-mentioned embodiment and the above-mentioned modified example, the learning period may or may not be included in the common parameters. The common parameters may be any setting values that can be commonly set for a plurality of models.

Furthermore, according to the modified condition diagnosis system 1, the display of the diagnosis results (reference model graph J11, time series graphs J12, J13) shows the individual parameters (first thresholds f1 to f3, second thresholds g1 to g3) obtained by applying the common parameters to the reference model. Therefore, the user can understand from the display of the diagnosis results what thresholds the generalized common parameters correspond to in the diagnosis of each monitored object 101.

Furthermore, according to the condition diagnosis system 1 of this embodiment, when the common parameters are changed, the positions of the first thresholds f1 to f3 and the second thresholds g1 to g3 in the display of the diagnosis results (reference model graph J11, time series graphs J12, J13) change in accordance with the change in the common parameters. Therefore, the user can change the common parameter settings while taking into consideration how the common parameters affect the diagnosis results.

The above describes the embodiments of the present invention. However, the present invention is not limited to the above embodiments. For example, in the above embodiment, an example was shown in which the displacement of vibration was adopted as the physical quantity of the monitored object measured for diagnosis, but the physical quantity measured for diagnosis may be, for example, the frequency of vibration, temperature, current if the monitored object is an electrical device, distortion if the monitored object is a machine with a load, or iron powder concentration in lubricating oil if the monitored object is a mechanism, or various physical quantities may be adopted or added. In addition, in the above embodiment, an example was shown in which the RMS and peak value of the measured physical quantity were adopted as feature quantities (statistics) calculated from the measured physical quantities in order to calculate individual parameters used in diagnosing the state of each monitored object 101. However, various statistics such as the minimum value, median, average value, and standard deviation of the measured physical quantity may be adopted as the feature quantity. In addition, the number of types of feature quantities is not limited to two, and may be three or more. In the above embodiment, rotating equipment such as a reducer and a motor is used as an example of the monitored object, but various elements such as various mechanical parts such as a chain sprocket and various electrical parts such as a control board may be applied as the monitored object 101. In addition, various systems such as an injection molding machine, a machine tool, and an industrial robot may be applied as a system to be diagnosed that includes multiple monitored objects. In the above embodiment, an example is shown in which the condition diagnosis system 1 has a condition diagnosis device 10 and a comprehensive management device 20, and processing is divided and executed by two systems, but a configuration in which one device executes all the processing, or three or more devices that operate in cooperation with each other via communication may share and execute the processing. In addition, the details shown in the embodiment can be changed as appropriate within the scope of the invention.

REFERENCE SIGNS LIST 1 Condition diagnosis system 10 Condition diagnosis device 11 Sensor signal acquisition unit 12 Communication processing unit 13 Memory processing unit 131, 132, 133 Storage unit 14 Arithmetic processing unit 141 AD conversion unit 142 Filter processing unit 143 Feature amount calculation unit 144 MD calculation unit 145 Reference model creation unit (determination unit)
146 Diagnosis unit 20 General management device 21 Common parameter setting processing unit (setting unit)
22 common parameter storage unit 23 operation processing unit 24 display processing unit 25 communication processing unit 26 notification processing unit 31 input device 32 display device 100 system to be diagnosed 101 monitored object 102 sensor MS1, MS2 reference model e1 to e3 boundary (individual parameters)
P0, P1 Plot H1 Setting screen L1-L3 Input field H2-H4 Diagnosis result screen J1, J11 Reference model graph J4-J6, J12, J13 Time series graph f1-f3 First threshold (individual parameter)
g1 to g3 Second threshold (individual parameters)

Claims (10)

A condition diagnosis system for diagnosing conditions of a plurality of monitoring targets, comprising:
A setting unit that sets a common parameter common to the plurality of monitoring targets;
a determination unit that determines a plurality of individual parameters to be used in diagnosing states of the plurality of monitoring targets, based on the common parameter and values of a plurality of sensors that measure physical quantities of the plurality of monitoring targets;
a diagnosis unit that diagnoses a state of each of the monitoring objects based on the individual parameters corresponding to the monitoring object and a measured value of the physical quantity of the monitoring object;
a display unit that displays a diagnosis result obtained by the diagnosis unit; and
Equipped with
A condition diagnosis system in which the diagnostic result is changeable by changing a common parameter in the setting unit while the diagnostic result is displayed on the display unit . the individual parameters are thresholds set for each of the plurality of monitoring targets, the thresholds being created from the common parameters and values of sensors provided for each of the plurality of monitoring targets;
The display unit displays the threshold value in a manner that makes the threshold value identifiable.
The condition diagnostic system according to claim 1 . The display of the diagnosis result includes information on the degree of deviation from a standard in the state of the monitored object determined using the individual parameters.
3. A condition diagnosis system according to claim 1 or 2. The display of the diagnosis result includes time-series information of the deviation degree.
The condition diagnostic system according to claim 3. The display of the diagnostic result includes information indicating the individual parameters.
The condition diagnosis system according to any one of claims 1 to 4. When the common parameter is changed while the diagnostic result is displayed on the display unit , the individual parameter shown in the display of the diagnostic result changes.
The condition diagnostic system according to claim 5. The sensor measures vibration.
The condition diagnosis system according to any one of claims 1 to 6. The determination unit is capable of redoing the determination process of the plurality of individual parameters.
The condition diagnosis system according to any one of claims 1 to 7. The determination unit determines the individual parameters using statistics calculated from the values of the sensors;
the statistics include only a value indicating an average amount of the physical quantity obtained from the value of the sensor and a peak value;
The condition diagnosis system according to any one of claims 1 to 8. The determination unit determines the individual parameters using statistics calculated from the values of the sensors;
the statistics include only a value indicating an average amount of the physical quantity obtained from the value of the sensor and a peak value;
The display of the diagnosis result includes displaying a graph in which one of the value indicating the average amount and the peak value is on the vertical axis and the other is on the horizontal axis.
The condition diagnosis system according to any one of claims 1 to 6.

Images