KR100199105B1 - Vibration monitoring device of rotating body with learning function - Google Patents

Abstract

The disclosed invention relates to a vibration diagnosis apparatus for a rotating body having a learning function for analyzing a vibration signal from a rotating body such as a motor, an engine, etc. to find out the cause of the abnormality, and a vibration signal measuring the vibration of the rotating body such as a rotating shaft. After signal processing, we can make a signal pattern using an artificial neural network and quickly determine the cause of the abnormality according to the similarity with the registered standard pattern, and predict the abnormality of the rotating body. Therefore, it is possible to stably operate the mechanical device including the rotating body, and it is possible to quickly find the cause of the abnormality suitable for the use environment, so the reliability of the diagnosis result is high.

Description

Vibration Diagnosis Device of Rotating Body with Learning Function

The present invention compares the standard pattern of the vibration signal possessed in advance with the signal pattern of the rotating body currently in operation and finds the cause of the abnormality. The unregistered signal pattern is registered as a new standard pattern, thereby increasing the reliability of the abnormal diagnosis. The present invention relates to a vibration diagnosis device for a rotating body having a learning function to accurately diagnose a cause.

In general, a mechanical device uses a motor, an engine, or the like as a power source, and a rotating shaft connected to the power source transmits power from the power source to a device for processing an object. It is noted that such a rotating shaft is widely applied to machinery. The rotating shaft that transmits power inevitably vibrates during the power transmission process, and if the vibration of the rotating shaft is at an acceptable level, the predetermined process can be performed without a problem. This is very important.

In addition, the vibration of the rotating body including the rotating shaft is generated not only by the power transmission process but also by the coupling state of the power source and the rotating shaft or other factors, and thus it is adopted as a criterion for finding the cause of abnormality. That is, the vibration generated in the process of transmitting power by the rotating shaft is amplified by the abnormality of the rotating body (unbalanced mass, axial deviation, shaft bending, etc.), rolling bearing abnormality, journal bearing abnormality, gear abnormality, loosening abnormality, etc. If the measurement falls within the allowable range of the normal value, it can be determined that the rotating body operates normally.

As such, a technique for determining abnormal diagnosis of a mechanical apparatus based on vibration of a rotating body is disclosed in US Pat. Nos. 4,437,163 and 44,48077. Conventionally, an instrument for measuring the vibration of the rotating body is installed, and abnormal diagnosis is performed based on the vibration signal measured by the instrument. That is, when the magnitude of the vibration signal is out of the normal value, it is determined as an abnormality factor and a user is notified of an alarm message. In this method, it was found that the rotating body malfunctioned when excessive vibration occurred, but it was not known which element of the mechanism was the problem. In another method, the expert system experimentally analyzes the vibration of the rotating body by factor and repeatedly inspects it, and by using the test data, it makes a diagnosis rule necessary to diagnose the factor and analyzes the vibration of the rotating body according to this diagnosis rule. By identifying abnormal factors. However, this method requires a lot of time for the expert to make a diagnosis rule, and there is a concern that subjective physicians are involved in the preparation of the diagnosis rule. In addition, in order to analyze the vibration of the rotating body in real time according to the diagnostic rule, there was a drawback that a large-capacity high-performance computer must be provided.

In the prior art, the magnitude of the vibration signal with respect to the normal value is determined through experience and experiment in advance, and when the vibration signal exceeding the magnitude is merely informed to the user of the abnormality, the specific abnormal cause could not be found. The failure of the mechanism could not be predicted in advance. In addition, it is possible to find the cause of an abnormality by using a diagnosis rule by an expert, but subjective factors of the expert may intervene, and there are limitations such as having to have a high-performance computer for performing a complicated algorithm to perform the abnormal diagnosis. Followed.

Accordingly, an object of the present invention is to analyze and maintain the standard pattern of the vibration signal for each abnormality factor and to compare the signal pattern of the vibration signal applied from the currently operating rotating body to find the cause of the abnormality, a new type signal pattern from the rotating body When it is applied to provide a learning function that registers as a standard pattern to provide a vibration diagnosis device for a rotating body having a learning function that can quickly diagnose the cause of abnormalities.

1 is a block diagram of an apparatus for diagnosing vibration of a rotating body according to the present invention;

2 is a view showing a processing operation of the pattern analysis unit of the present invention.

* Description of the symbols for the main parts of the drawings *

10: signal processing unit 11: pattern analysis unit

12: standard pattern setting unit 13: pattern reset unit

14: storage unit 15: error recognition unit

16 interface unit

An object of the present invention as described above in the apparatus for diagnosing the cause of the abnormality based on the vibration of the rotating body, receiving the vibration signal of the rotating body detected from the rotating body, the vibration signal in the time domain and frequency domain A signal processor for analyzing and outputting a discrete vibration signal having a discrete magnitude level; An artificial neural network for generating a standard pattern based on a test discrete vibration signal input to the signal processing unit, and a weight factor is set for each node of the input layer, hidden layer, and output layer constituting the artificial neural network. A pattern analysis unit for comparing a signal pattern corresponding to the discrete vibration signal currently applied to the current signal processing unit with a pre-registered standard pattern and outputting a similar information signal indicating a similar degree of the two patterns; A storage unit for storing data and additional data about standard patterns corresponding to abnormal causes from the pattern analysis unit; A standard pattern setting unit which selects and outputs a similar standard pattern that is most similar to pre-registered standard patterns corresponding to a current discrete vibration signal from the pattern analysis unit; A pattern reset unit configured to reset and output a pseudo standard pattern input from the standard pattern setter in accordance with a current discrete vibration signal; And a table in which the similar information signal and the abnormal cause correspond to each other, and are determined by an error recognizing unit which determines an abnormal cause according to the similar information signal received from the pattern analysis unit.

Hereinafter, a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an apparatus for diagnosing vibration of a rotating body according to the present invention. In FIG. 1, the signal processing unit 10 is installed in a rotating body and receives a predetermined vibration signal from a measuring instrument measuring a change caused by vibration. The signal processor 10 generates a discrete vibration signal represented in a discrete form by analyzing the signal pattern in the time domain and the frequency domain, respectively, so as to receive the analog signal and analyze the signal pattern. The storage unit 14 receives and holds standard patterns corresponding to abnormal causes from the pattern analysis unit 11, and outputs the corresponding standard patterns to the pattern analysis unit 11 when an output from the pattern analysis unit 11 is requested. do. The standard pattern setting unit 12 receives information about a standard pattern registered in advance from the pattern analysis unit 11 and outputs a similar standard pattern. The pattern reset unit 13 receives a standard pattern corresponding to the current discrete vibration signal by receiving the most similar standard pattern from the standard pattern setting unit 12 when generating a new standard pattern based on the current discrete vibration signal. It resets and outputs the reset signal pattern to the pattern analysis part 11. As shown in FIG. The pattern analysis unit 11 receives a signal pattern for the current discrete vibration signal from the pattern reset unit 13, and compares the applied signal pattern with a pre-registered standard pattern to indicate a similar degree of the two patterns. The pseudo information signal is output to the error recognition unit 15. The error recognition unit 15 finds the cause of the abnormality on the table based on the similar information signal, and holds a table corresponding to the similar information signal for each abnormal cause in advance. If the error recognition unit 15 outputs predetermined information on the cause of the abnormality found on the table to the interface unit 16, the user can check the vibration factor of the current rotating body through the interface unit 16.

The pattern analysis unit 11 has a learning function, and may additionally register a new standard pattern, that is, a standard pattern obtained by the discrete vibration signal applied from the current signal processing unit 10, to a pre-registered standard pattern. To help understanding, the operation of the pattern analysis unit 11 will be described with reference to FIG. 2.

FIG. 2 is a diagram illustrating a processing operation of the pattern analysis unit of the present invention, and schematically illustrates a process in which the pattern analysis unit 11 receives a discrete vibration signal and generates a similar information signal. Here, the pattern analysis unit 11 applies a 'back propagation network model' algorithm, which is a kind of a known artificial neural network. In FIG. 2, the discrete vibration signal has a magnitude level of 0 to 1 and appears in a discrete distribution form. The pattern analysis unit 11 receives the discrete vibration signal, processes it according to the back propagation, and outputs a similar information signal to the error recognition unit 15.

The pattern analysis unit 11 is composed of three layers constituting an artificial neural network, that is, an input layer, a hidden layer, and an output layer. Each layer performs arithmetic processing according to the weight factor given to the nodes indicated by the circle. The weight factor of the nodes of each layer may be adjusted by the user, and the weight factor may be changed through the interface unit 16 described above. The input layer multiplies the discrete vibration signal received by the weight factor of each node and sends it to the node of the hidden layer, and the node of the hidden layer multiplies the weight factor and then sends it to the output layer. .

The pattern analysis unit 11 adjusts the weight factor of each layer until the error of the similar information signal obtained by the calculation process in each layer is within the allowable range, and adjusts the similar information signal generated by the final adjusted weight factor. Output to error recognition part. Then, the error recognition unit 15 finds an abnormal cause corresponding to the similar information signal on the table and informs the user through the interface unit 16.

According to the present invention as described above, after processing the vibration signal measuring the vibration of the rotating body such as a rotating shaft to create a signal pattern using an artificial neural network and quickly determine the cause of the abnormality according to the similarity with the registered standard pattern. In addition, it is possible to predict whether there is an abnormality in the rotating body which is currently operating, so that the machine can be stably operated. In addition, according to the present invention, a new standard pattern can be newly registered by the pattern recognition unit that analyzes the pattern of the vibration signal, thereby making it easy to find the cause of the abnormality suitable for the use environment and improve the reliability of the diagnosis result of the rotating body. It can be effected.

Claims (3)

In the device for diagnosing the cause of abnormality based on the vibration of the rotating body, A signal processor for receiving a vibration signal of the rotating body detected from the rotating body and analyzing the vibration signal in a time domain and a frequency domain to output a discrete vibration signal having a discrete magnitude level; An artificial neural network for generating a standard pattern based on a test discrete vibration signal input to the signal processing unit, and a weight factor is set for each node of the input layer, hidden layer, and output layer constituting the artificial neural network. A pattern analysis unit for comparing a signal pattern corresponding to the discrete vibration signal currently applied to the current signal processing unit with a pre-registered standard pattern and outputting a similar information signal indicating a similar degree of the two patterns; A storage unit for storing data and additional data of standard patterns corresponding to abnormal causes from the pattern analysis unit; A standard pattern setting unit which selects and outputs a similar standard pattern that is most similar to pre-registered standard patterns corresponding to a current discrete vibration signal from the pattern analysis unit; A pseudo standard pattern received from the standard pattern setting unit is applied to a current discrete vibration signal. A pattern reset unit which resets and outputs the matching unit; And Vibration of the rotating body having a learning function characterized in that it comprises a table corresponding to the similar information signal and the cause of error, and including an error recognition unit for determining the cause of the abnormality in accordance with the similar information signal received from the pattern analysis unit Diagnostic device. The method of claim 1, wherein the pattern analysis unit adjusts the weight factor of each layer until the error of the similar information signal obtained by the calculation process in each layer is within an allowable range, and according to the final adjusted weight factor. Vibration diagnostic apparatus of a rotating body having a learning function, characterized in that for making. The apparatus of claim 1, wherein the pattern analysis unit adjusts a weight factor applied to a node of each layer according to an input command by a user.