KR19990066119A - Error Diagnosis Method of Rotating Machine Using Fuzzy Logic - Google Patents

Abstract

The fault diagnosis method of a rotating machine using fuzzy logic is to monitor the abnormal operation signal based on the vibration signal measured from the rotating shaft element of the mechanical device, and to diagnose the cause of the fault using the fuzzy reasoning technique. The present invention is a vibration information input step of inputting the signals detected from a plurality of displacement sensors and acceleration sensors installed on the rotational drive elements, the rotational speed input in the input step and the vibration signal and installation information of each sensor as input factors Based on the first diagnosis step of presenting the individual diagnosis result of each sensor through fuzzy inference, and based on the expert's knowledge, the weighting factor considering the frequency range of each defect, the number of input sensors and the type of sensor is applied. A second diagnostic step is included that suggests the final cause of anomalies from the individual results of the first diagnosis. Accordingly, the present invention monitors and diagnoses abnormal driving symptoms and warns the driver of various abnormal causes that may occur during operation, thereby preventing the failure of the machine in advance, and estimating the replacement time of the rotating body components, thereby causing a fatal failure. Previously, the operator can repair the device, which not only increases work efficiency but also improves the reliability of the rotary drive system.

Description

Abnormal Diagnosis Method of Rotating Machine Using Fuzzy Logic

The present invention relates to a vibration diagnosis system for diagnosing abnormal conditions on the basis of vibrations of a rotating machine, and more particularly, to monitor abnormal driving signs based on vibration signals measured from rotating shaft elements of a mechanical device. The present invention relates to a method for diagnosing abnormalities of a rotating machine using fuzzy logic that enables reliable diagnosis using fuzzy reasoning technique.

In general, a machine converts a rotational motion of a power source such as a motor or engine into a motion type suitable for the purpose of the device. Therefore, a rotation transmission device such as a rotating shaft for transmitting the motion of the power source is an essential element in the machine. Rotating elements such as the rotating shaft is not only a vibration light occurs in accordance with the rotational movement but also a friction light is formed to act as a factor causing an abnormal state. In order to solve this problem, most rotary machines have constructed a system for self-diagnosis by identifying the vibration state of the rotating shaft. The above-mentioned vibration diagnosis system which is currently used can be divided into two types. One system sets the limit size of the vibration signal through experience and experiment in advance, and provides the driver with a vibration signal exceeding the magnitude. I am using a method to warn of an abnormal condition. Such a system can be used to determine whether an object is abnormal, but it is not possible to proactively cope with the operation abnormality of the target system in advance, and it is not possible to know the specific cause of the error. There has been the hassle of additional cause detection.

In addition, the other diagnosis system is to find out the cause of the abnormality by performing a search for the diagnosis rule provided by the expert in advance, and it takes a lot of time to prepare the diagnosis rule, subjective individual's subjective There may be a difference, and a high performance computer is required to perform a vibration diagnosis by searching a relation tree in real time. Moreover, it is fundamentally impossible to draw up diagnostic rules for abnormalities in all cases, so in the case of signals with indefinite patterns such as vibration signals, there is a possibility that diagnosis can not be made due to missing diagnostic rules. .

Accordingly, an object of the present invention is to devise a solution to the above-mentioned conventional defects, and to monitor the abnormal driving signal based on the vibration signal measured from the rotating shaft element of the mechanical device, and to purge the cause of the abnormality. Rotating machine using fuzzy logic that diagnoses through an inference technique and warns the driver of various abnormalities that may occur during operation, and prevents the failure of the machine in advance and enables the driver to repair the device before the fatal failure occurs. To provide a diagnostic method for abnormalities.

1 is a configuration diagram showing a vibration machine vibration diagnosis system to which the present invention is applied,

2 is an internal structural diagram for diagnosing by fuzzy inference of the primary diagnosis of FIG.

3 is a view for explaining the diagnosis by the fuzzy expert of FIG.

Figure 4a-4d is a graph showing the membership function of the displacement sensor installed in the journal bearing applied to an embodiment of the present invention.

* Explanation of symbols on the main parts of the drawings

10: rotating machine 20: vibration signal input unit

21 sensor 22 signal processor

30: 1st diagnosis part 40: 2nd diagnosis part

The abnormal diagnosis method of a rotary machine using the fuzzy logic of the present invention for achieving the above object, in the abnormal diagnosis method of the rotary drive system, the signals detected from a plurality of displacement sensors and acceleration sensors installed on the rotary drive element Inputting vibration information; A first diagnosis step of performing fuzzy inference based on the rotational speed inputted in the input step, the vibration signal of each sensor, and installation information as input factors to present an individual diagnosis result of each sensor; And a second diagnosis step that proposes the final cause of abnormality from the individual results of the first diagnosis by applying weights in consideration of the frequency range of each defect, the number of input sensors, and the type of sensor based on expert knowledge. It includes.

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

1 shows a configuration diagram of a rotary machine vibration diagnosis system to which the present invention is applied. The system shown in FIG. 1 includes a rotary machine 10, which is an object of abnormal diagnosis, and a vibration signal input unit 20 for measuring a vibration signal from a rotary shaft element of the rotary machine 10 and inputting it as a signal for diagnosis. . The vibration signal input unit 20 is connected to the plurality of sensors 21 installed on the rotary drive elements of the rotary machine 10 and the sensors 21 correspondingly, and the vibration measured by the sensors 21. It consists of a plurality of signal processors 22 which process a signal. In addition, the primary diagnosis unit 30 using the fuzzy inference technique, the secondary diagnosis unit 40 for the final diagnosis by combining the primary diagnosis results of the primary diagnosis unit 30, and the secondary diagnosis unit 40 A user interface (GUI) 50 provided with the final diagnosis result is provided. An operation of the system of FIG. 1 having such a configuration will be described in more detail with reference to FIGS. 2 to 4.

The sensors 21 of the vibration signal input unit 20 measure vibrations of the rotational driving elements of the rotary machine 10, which is an object of abnormal diagnosis. One of the sensors 21 measures the speed of rotation using a photocoupler. The signal processors 22 process a predetermined signal so as to recognize the signal measured by the corresponding sensors 21 in the primary diagnosis unit 30 and input the signal to the primary diagnosis unit 30. The primary diagnosis unit 30 performs fuzzy inference based on vibration signals, rotational speeds, and sensor setup information, that is, installation information of the sensors 21, and presents individual diagnostic results of each sensor. . Here, the installation information used includes an installation direction, a type of sensor, a measurement position, and the like. In addition, the secondary diagnosis unit 40 individually applies the weights in consideration of the frequency domain of each defect, the number of input sensors, the type of the sensor, and the like, based on the expert's knowledge. The final cause of the problem is suggested. This will be described in more detail below.

1. Input of vibration signal

The system is equipped with a plurality of sensors 21 to obtain the state of the rotating shaft of the rotary machine 10, of which the displacement sensor can directly measure the displacement of the rotating shaft to grasp information about the behavior of the rotating shaft. The acceleration sensor is attached to the shaft support or base and used to obtain information about the bearings and housing that support the shaft. In particular, in this system, an appropriate number is installed to receive up to four signals for each of the acceleration sensor and the displacement sensor, which is random in the process of synthesizing the first diagnosis result of each sensor signal during the second diagnosis. This is because it is not possible to database all the weights for the combination of the first order diagnostic results of a number of sensors.

In addition, the rotational shaft changes the vibration force of the whole system by changing the excitation force acting on the shaft and the rigidity supporting the shaft as the rotational speed changes, so the rotational speed is taken as a reference signal for diagnosis.

2. Diagnosis

First diagnosis is a step of performing a fuzzy inference, Figure 2 is an internal structural diagram for the diagnosis by the fuzzy inference of the first diagnostic unit 30, Figure 3 is a fuzzy expert of the primary diagnostic unit 30 It is a figure for demonstrating diagnosis. With reference to these two drawings will be described in detail the first diagnostic process.

In general, fuzzy logic is more complex than the existing quantitative methods, and the information obtained is qualitative, inaccurate, and uncertain. have. Therefore, the application is suitable when a long time experience of an expert is required, such as the diagnosis of abnormal causes that appear in combination with various causes such as vibration of a rotating shaft. Now, the first diagnosis process will be described in more detail in order.

Normalization

The input variable of the fuzzy diagnosing machine is information about the system state which the expert refers to to diagnose the defect of the rotating body, and it is the state value of the rotating body measured by the rotational speed, displacement sensor and acceleration sensor. It becomes a defect of. Since the characteristics of the vibration signal of the rotating shaft can be easily observed through the power change of a specific frequency in the frequency domain rather than in the time domain, the sensor signal is decomposed into each frequency component by A / D conversion and then FFT processed, and then the defect of the rotating body The values of frequency components related to are used as input variables. Table 1 below shows the types and definitions of these input variables, and Table 2 shows the types and definitions of fuzzy output variables as a result of the diagnosis to be obtained from the fuzzy diagnostics.

Type and Definition of Input Variables No Symbol variable Justice One SP V Rotational Speed of Rotating Body (rpm) 2 MAG M _1x Amplitude at Frequency of Rotational Speed (1X) 3 Q1 M _Q1 Amplitude Average at 1-40% Frequency of 1X 4 0 W M _0W Maximum amplitude at 40 to 50% frequency of 1X 5 Q2 M _Q2 Amplitude mean at 50 to 60% frequency of 1X 6 2X M _2X Amplitude of 2X frequency band 7 3X M _3X Amplitude of 3X frequency band 8 4X M _4X Amplitude of 4X frequency band 9 B1 M _B1 Amplitude at Ball Bearing Fault Frequency 1 10 B2 M _B2 Amplitude at Ball Bearing Fault Frequency 2

Output Variable Types and Definitions No Symbol variable Justice One MA D _MA Misalignment 2 UB D _UB Unbalance 3 LS D _LS Loosness 4 BD D _BD Ball bearing defect 5 RB D _RB Rubbing 6 OW D _OW Oil whirl 7 SHI D _SHI Subharmonic instability

Since the rotational speed of the rotor has a wide range of values depending on the application, it is not suitable to use the absolute value directly as an input variable of the fuzzy inference machine. Therefore, in the present invention, the relative ratio with the current operating rotational speed (V) by receiving the maximum rotational speed (V _max ) of the target rotating body system input variable ( )

In addition, since 1X component is a main component of rotating body vibration and a component synchronizing with the rotational speed, its magnitude becomes an important criterion for diagnosis. However, since the absolute magnitude of the system varies depending on the system configuration and operating conditions, the absolute value cannot be directly converted into an input variable. Therefore, in the present invention, the amplitude value (M _warning ) corresponding to the warning level of the target system is input and the input variable of the fuzzy inference unit as a state ratio based on the value ( )

In the case of the input variables except SP and MAG in Table 1, the ratio and shape of the relative size of the rotor rather than the absolute value of the input give much information to the diagnosis. Thus, Q1,0W, Q2,2X, 3X, 4X , B1, B2 may be converted as shown below as a percentage of the amplitude M _1x in 1X.

i = Q1,0W, Q2,2X, 3X, 4X, B1, B2

-Fuzzyfication of input / output variables

Input variables are expressed in linguistic form according to their values for the linguistic representation of the fuzzy inference rule. According to the present invention, the fuzzy division function having a triangular shape is used to purge each input space. At this time, the fuzzy partitioning function for each input space has five linguistic values (ZE, SM, ME, LA, VL), and the number of partitions varies depending on the input variable. Most input variables are divided into four membership functions, with five membership functions for variables that need to be examined more closely, and three membership functions for small influences.

The center value of each fuzzy control function is obtained based on the expert's knowledge of the vibration diagnosis of the rotating body. When it is changed into linguistic expression, it represents the linguistic value, and the same input variable may have different values depending on the situation. Can be. In particular, since the response characteristics for the same frequency components are different depending on the type of measurement sensor, that is, the displacement sensor or the acceleration sensor, the magnitude of the center value must be different. It will also change depending on where it is installed. As an example, Figure 4 shows a graph of the membership function of the displacement sensor installed in the journal bearing applied to an embodiment of the present invention.

-Preparation of Fuzzy Reasoning Rules for Diagnosis

In principle, all ten input variables shown in Table 1 should be considered simultaneously. In that case, however, the number of fuzzy inference rules required becomes too large. Therefore, in the present embodiment, in the fuzzy inference rule for the displacement sensor as defined in Tables 3a and 3b below, the MA from the superharmonic components 2X, 3X, 4X, B1, B2 based on the 1X component is used. The output result for, UB, LS, BD is obtained, and the output result for RB, 0W, SHI is obtained from the subharmonic components Q1,0W, Q2. And, inference rules of displacement sensor and acceleration sensor are distinguished. This is because, as described above, the two sensors are different in response to the abnormal signal. In particular, the present invention uses the Min-Max fuzzy inference method. Since this method is a universal method of fuzzy inference, a description thereof will be omitted.

Super harmonic SP MAG 2X 3X 4X B1 B2 MA UB LS BD R1 SM ME ZE LA ME ZE ZE SM ME SM ZE R2 SM ME ME ME ME ME ME ZE SM ZE SM R3 SM ME ME LA ME ME ME ZE SM SM ME · · ·

Subharmonic SP MAG Q1 0 W Q2 RB 0 W SHI R1 SM SM LA SM VL ZE ZE SM R2 SM ME ME LA SM SM SM SM R3 SM ME ME VL ME SM ME ZE · · ·

3. Secondary diagnosis

Most causes of rotor failure can be inferred by separating the frequency domain. However, in the case of a specific abnormal cause, it may have a significant influence on a frequency region other than the specific frequency. In such a case, the weight is set on the cause of the inferred subharmonic region and the cause of the super harmonic region. To give. Tables 4a to 4c show weights for the multiple displacement sensors.

When there are two sensors One 2 MA UB LS MA BA 0.8 LS BA 0.7 UB 0 W 0.5

When there are three sensors One 2 3 UB RB SUB MA MA UB 0.7 UB UB 0 W 0.8 0 W SUB SUB 0.8

4 sensors number 1 No.2 number 3 4 times MA LS RB MA MA MA 0 W 0.8 LS LS LS BA 0.8 RB RB 0 W 0 W 0.7

As shown in Tables 4a to 4c above for the individual diagnosis results of the respective sensors that have undergone the above processes, the results are summarized for each sensor type in the displacement sensor and the acceleration sensor. As the number of sensors increases, the reliability of the judgment is adjusted by weighting each combination of the results inferred from each sensor so as to increase the reliability of the judgment. For example, if the oil wheel is inferred from one input of four displacement sensors and the axis misalignment is inferred from the other three sensors, the probability of shaft misalignment is adjusted by considering the influence of the oil wheel's overall system. .

In addition, it is finally necessary to compare the diagnosis result obtained from the displacement sensors with the diagnosis result obtained from the acceleration sensors. For example, the oil wheel (0W) is easy to detect through a displacement sensor that measures the direct movement of the shaft. In the case of mechanical loosening (LS), the acceleration sensor installed in the housing is advantageous. Inevitably, the final diagnosis is made by applying weights as shown in Table 5 below.

Weight Considering Sensor Characteristics Displacement acceleration Displacement sensor Acceleration sensor MA UB MA UB MA LS 0.8 UB LS 0.8 UB MA 0.8 MA UB 0.8

The weight of the secondary diagnosis described above is prepared by an expert on the rotational fault diagnosis, and appropriate modification is necessary to suit the characteristic in consideration of the characteristics of the applied system.

As described above, the fault diagnosis method of the rotating machine using the fuzzy logic of the present invention monitors and diagnoses the abnormal operation symptoms on the basis of the vibration signal measured from the rotating shaft element of the mechanical device to identify various causes that may occur during operation. By alerting the operator to prevent the failure of the machine in advance, and to estimate the replacement time of the rotor components, the operator can repair the device before the fatal failure. There is an effect that can increase the reliability.

Claims (5)

In the abnormal diagnosis method of the rotation drive system, A vibration information input step of inputting signals detected from a plurality of displacement sensors and acceleration sensors installed on the rotation driving elements; A first diagnosis step of performing fuzzy inference based on the rotational speed inputted in the input step, the vibration signal of each sensor, and installation information as input factors to present an individual diagnosis result of each sensor; And Based on the expert's knowledge, the second diagnosis step is proposed to propose the final cause of abnormality from the individual results of the first diagnosis by applying the weights in consideration of the frequency range of each defect, the number of input sensors, and the types of sensors. An abnormal diagnosis method of a rotating machine using a fuzzy logic. The method of claim 1, wherein the installation information used in the first diagnostic step includes an installation direction, a type of sensor, a measurement position, and the like. The method of claim 1, wherein the first diagnostic step Normalizing the state values of the rotating body measured by the rotational speed, the displacement sensor and the acceleration sensor to an input variable suitable for fuzzy inference; A fuzzy step of expressing the normalized input variable in a linguistic form suitable for a fuzzy inference rule; Performing fuzzy reasoning on the fuzzy input variable by applying fuzzy reasoning rules; And And diffusing the fuzzy inferred values and outputting the fuzzy inferred values. The method of claim 3, wherein the rotational speed of the input variable in the normalization step ( ) Is calculated as the ratio between the maximum rotational speed (V _max ) and the current operating speed (V) of the target rotor system, and 1X component ( ) Is calculated as a relative ratio based on the amplitude magnitude (M _warning ) corresponding to the warning level of the target system, and other input variables ( ) Is converted as a ratio of the amplitude (M _1x ) at 1X and normalized as follows. next , here, i = Q1,0W, Q2,2X, 3X, 4X, B1, B2 4. The method of claim 3, wherein the fuzzy reasoning step uses a Min-Max method.