KR101165078B1 - Fault Diagnosis Apparatus for Press Process and method at the same - Google Patents

Abstract

The present invention relates to a failure diagnosis apparatus and a method of a pressing process.

To this end, the present invention is a vibration sensor for detecting a vibration signal in the electric motor, a data collector for digitally storing the vibration signal sensed by the vibration sensor and a frequency domain for the vibration signal data stored in the data collector Provided is a fault diagnosis apparatus comprising a frequency analyzer for converting, and a fault classifier for classifying the types of faults generated in the motor by using the vibration signal data converted through the frequency analyzer as an input pattern, and a fault diagnosis method using the same. .

Press Process, Induction Motor, Fault Diagnosis, FFT

Description

Fault Diagnosis Apparatus for Press Process and method at the same}

The present invention relates to a failure diagnosis apparatus of a product, and more particularly, to a failure diagnosis apparatus and a method in a press process.

Press processing, which is widely used from precision parts of watches and cameras to automobile bodies, should be operated continuously for 24 hours. Minor breakdowns in the press process are now immediately handled by the field maintenance team. In the case of large failures that cannot be dealt with in advance, such as main motor failures, external repair companies usually repair the failures by replacing some parts. In this case, the press line operation cannot proceed and the entire plant is stopped due to the failure to supply the steel sheet material required for the welding line in a timely manner. For example, according to data from a press company, the downtime of the press processing line in 2007 was 204.27 hours, accounting for about 40% of the entire plant, and the loss of production from press non-operation was about 7 million won. In addition, the amount of preservation loss was estimated at 4.5 million, and the press outsourcing repair occurred 8 times, resulting in repair cost of about 30 million won. Among them, there were many bearing replacements, and the total cost was about 41 million won in terms of losses due to press downtime.

As such, if the press line work cannot be carried out, the cost is considered to be enormous if it is converted to the entire press line currently operating nationwide if the price is 41 million won per unit. The phenomenon of stopping occurs.

In order to prevent the failure of the current press process, there is a method of checking whether there is an abnormality of the main motor, the maintenance officer directly goes to the upper end of the press to check the operation sound of the main motor. However, it is never easy for the person in charge to inspect the inspection areas located above and below the press every day, and it is difficult to make accurate judgments because it reflects the subjective opinions of the person rather than the objective data, and there are problems of safety accidents such as the possibility of the fall of the conservation person. have.

In addition, in a mechanical press, a certain amount of lubricant must be continuously flowed to reduce friction between the bearing and each shaft. That is, the hydraulic pump is pulled out of the lubricating oil tank and delivered to the end bearings and shafts through each ripple through the hose, and the lubricating oil tank is checked daily for smooth supply of the lubricating oil. Constant flow must be maintained. However, if the lubricating oil supply is not smooth due to missed check period, the amount of lubricating oil supplied through the ripple is small, so that the friction of the bearings of each shaft may increase, or noise and heat may occur, and the shaft may melt due to heat. In this case, each ripple, oil filter, and oil tank of the press must be replaced, and the cost of the loss can be up to tens of millions of won.

As described above, since the loss cost due to the failure of the press process line is enormous, there is an urgent need for a failure diagnosis system that can efficiently diagnose the problem and notify a person in charge.

In addition, it is critically required to establish a lubricator automatic inspection system that can automatically check the state of the lubricator so as to maintain a constant amount of lubricating oil at all times and inform the person in charge of status information.

The present invention has been made to solve the above problems, and an object of the present invention is to provide a failure diagnosis system that can efficiently diagnose a failure occurring in a press process line and notify a person in charge.

In addition, it is an object of the present invention to provide an automatic lubricator inspection system that can automatically check the state of the lubricator to maintain a constant amount of lubricating oil, and inform the person in charge of the status information.

In order to achieve the above object, a failure diagnosis apparatus according to a preferred embodiment of the present invention includes a vibration sensor for detecting a vibration signal in an electric motor, a data collection unit for storing digital data of the vibration signal detected by the vibration sensor; A frequency analyzer for converting the vibration signal data stored in the data collector into a frequency domain, and a fault classifier for classifying the types of failures generated in the motor by using the vibration signal data converted through the frequency analyzer as an input pattern. It is characterized by including.

In addition, the fault classifier is characterized by classifying the type of fault using the ART2 neural network having a number of boundary factors.

In addition, the frequency analysis unit FFT converts the vibration signal data into a frequency domain, divides the FFT transformed data into each 100Hz frequency unit, and extracts one signal having a maximum amplitude value from the separated 100Hz unit. It is characterized in that for performing the monitoring frequency band.

The apparatus may further include a lubricating oil level sensor for measuring a lubricating oil level of the lubricating oil tank, and a lubricating oil level determining unit for determining whether or not a failure is compared with a reference value with respect to the level of the lubricating oil measured by the lubricating oil level sensor.

According to another preferred embodiment of the present invention, a failure diagnosis method includes detecting a vibration signal in an electric motor, storing the vibration signal as digital data, converting the vibration signal data into a frequency domain, and converting the And classifying a type of failure generated in the motor by using the vibration signal data as an input pattern.

In addition, the fault classification step is characterized by classifying the types of faults by using ART2 neural network having several boundary factors.

In the frequency domain conversion step, the FFT transforms the vibration signal data into a frequency domain, divides the FFT transformed data into each 100 Hz frequency unit, and extracts one signal having a maximum amplitude value from the separated 100 Hz unit. And performing the method for the monitoring frequency band.

The method may further include automatically checking a lubricant level by automatically detecting a lubricant level of the lubricant tank to determine whether there is a failure.

The present invention, by automating the failure diagnosis of the existing manual method in the press process, can enjoy the effect of cost reduction, efficiency improvement, safety accident prevention.

In addition, if the ART2 neural network according to the present invention is used as a failure classifier in a nonlinear system such as a press process, it is possible to actively cope with the occurrence of an unexpected new failure.

Hereinafter, a failure diagnosis apparatus and a method in a press process line according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a functional block diagram showing the configuration of the pressing process failure diagnosis apparatus 100 according to a preferred embodiment of the present invention, Figure 2 shows a failure diagnosis process of the press process through the failure diagnosis apparatus 100 of FIG. It is a flow chart.

As shown in FIG. 1, the failure diagnosis of the pressing process can be largely classified into a failure diagnosis for an induction motor which is a main motor of the pressing process, and an automatic inspection of a lubricant state in the pressing process.

First, looking at the failure diagnosis device 155 of the induction motor is configured as follows. That is, the vibration sensor 120 for detecting the vibration signal in the induction motor 110, the data collection unit 130 for digitally storing the vibration signal detected by the vibration sensor and the data collection unit 130 The frequency analysis unit 140 converts the stored vibration signal data into the frequency domain through FFT (Fast Fourier Transform), and the vibration signal data converted through the frequency analyzer 140 as an input pattern. The fault classifier 150 classifies the type of fault generated in the induction motor 110 through the ART2 neural network.

Here, the frequency analyzer 140 performs FFT conversion on the vibration signal data of the induction motor 110 detected through the vibration sensor 120. At this time, the frequency analyzer 140 divides the FFT-converted data into each 100Hz frequency unit, and extracts one signal having a maximum amplitude value from the separated 100Hz unit. In this way, a method of extracting one maximum value for each 100 Hz unit is performed for the monitoring frequency band. At this time, the monitoring frequency band is a frequency range set for the failure diagnosis may vary depending on the motor (motor) specification of the pressing process.

 In addition, the ART2 neural network learns in the same class if it is similar through a boundary factor test for checking the degree of agreement between the already learned pattern and the vibration signal data of the induction motor 110 which is the input pattern. There is an advantage that further learning is possible by classifying the failure pattern by the incremental classification algorithm that generates the class. Therefore, if the ART2 neural network is used as a fault classifier, it can effectively classify when an unexpected new fault occurs. Therefore, in the present invention, the ART2 neural network having various boundary factors, which complements the disadvantages of the existing ART2 neural network using one boundary argument, is used as the fault classifier 150 for the fault classification of the induction motor.

On the other hand, looking at the automatic lubricating oil inspection device 195 in the press process is configured as follows. That is, the lubricating oil level sensor 180 for measuring the lubricating oil level of the lubricating oil tank 170 and the level of the lubricating oil measured by the lubricating oil level sensor 180 are turned on / off compared with a reference value. It consists of a lubricant level determination unit 190 for determining whether or not the failure.

Here, the lubricant level sensor 180 is required to select the appropriate level sensor since the lubricant tank size currently used in the pressing process is 900mm * 600mm * 900mm, and the measurement range is 8 ~ when considering the distance and the position of the lubricant tank. It should be about 10mm. In particular, in consideration of distance and position of the lubricating oil tank, an ultrasonic level sensor in which the sensor unit and the transmitter display unit are separated is preferable.

As described above, the failure diagnosis device 155 of the induction motor in the press process measures the vibration value using a vibration sensor, and then classifies the failure (steady state, rotor failure, bearing using frequency analysis technique (FFT) and ART2 neural network). Fault) to diagnose the fault. In addition, the automatic lubricating oil inspection device 195 performs a lubricating oil level warning function for the ultrasonic level sensor.

Hereinafter, a press process failure diagnosis process according to the press process failure diagnosis apparatus 100 described with reference to FIG. 1 will be described with reference to FIG. 2.

First, the vibration signal value is sensed through the vibration sensor 120 in the induction motor 110 of the pressing process (S210).

Thereafter, the sensed vibration signal value is digitalized, sampled, and stored through the data collector 130 (S220).

Thereafter, the sampled vibration signal data is converted into a frequency domain value through the FFT method in the frequency analyzer 140 (S230). At this time, as shown in Figure 3a, the converted data is divided in units of 100Hz and the maximum amplitude value is extracted from the divided units. Thereafter, as shown in FIG. 3B, a total of 50 data are extracted in the range of 0 to 5 kHz, which is a monitoring frequency band.

Thereafter, the vibration signal data converted into the frequency domain is input to the fault classifier as an input pattern (S240).

Thereafter, the type of the failure is classified according to ART2 neural network using the converted vibration signal data as an input pattern, and the failure of the induction motor is diagnosed according to the classified failure (S250). In this case, as shown in FIG. 4, data to be applied to ART2 is again applied among the maximum amplitude value data extracted from the monitoring frequency band 1 (200 Hz to 600 Hz) and the monitoring frequency band 2 (1.5 KHz to 2.5 KHz) designated as the measurement frequency. When the selected data is input to ART2, the internal processing is used to determine whether there is a failure and to notify the outside. According to the result measured in ART2, a warning lamp is displayed according to normal, bearing failure and rotor failure in the troubleshooting part.

On the other hand, the ultrasonic level sensor 180 detects the lubricant level of the lubricant tank (S260), compares the detected lubricant level with a reference value to determine whether the lubricant is insufficient (S270), if it is insufficient diagnosed as a failure (S282) Otherwise, it is diagnosed as a normal state (S281) and displayed on a warning lamp. In this case, the reference value refers to a range of suitable lubricating oil levels according to the pressing process, and can be programmed and updated.

Hereinafter, a fault classifier using the ART2 neural network according to the present invention will be described in detail.

5 is a structural diagram of the fault classifier 150 using the ART2 neural network according to the present invention.

As shown in FIG. 5, the ART2 neural network responds to an input similar to a stored pattern to block layer 1 when a layer 1 for learning the pattern and a pattern not similar to the stored pattern are input, and a new class. It consists of Layer 2, which generates. Layer 1 consists of a two-layer neural network with an input layer and an output layer, and the input layer and output layer nodes are connected by weights in which learning patterns are stored. When the first input pattern is delivered to the ART2 neural network, since there is no generated class in the output layer, it is classified as the first class and the input pattern is stored as a weight between the first output node and the input nodes. The next time another pattern is entered, the similarity with the first class is checked to learn from the same class or create a new class.

The similarity test is determined by checking the boundary argument condition. The boundary argument is a value that determines whether two patterns are learned in the same class or classified as a new class. Here, the fault classifier using the ART2 neural network used in the fault diagnosis apparatus according to the present invention is characterized by having a number of boundary factors. This is to ensure that even if an unexpected new fault occurs, it can be classified.

6 is a flowchart summarizing the operation of the fault classifier 150 using the ART2 neural network in the fault diagnosis apparatus according to the present invention.

The structure of the ART2 neural network is the same as that of the existing ART-2 neural network, as shown in FIG. 5, but by using one boundary factor when classifying patterns in the existing ART2 neural network, New distance measurement and boundary factor conditions are used to compensate for the difficulty of considering sensitivity differences. That is, when a new pattern is input to the neural network, the distance between the input pattern and each output node is calculated as in Equation 1 below.

는 가중치가 부여된 무한대 노옴이며, E = diag(1/ε₁, 1/ε₂, ....1/ε_N) 로서 N×N 차원의 대각 가중치 행렬이고, ε_i 는 i번째 입력노드의 입력패턴에 대한 경계인수이며 ｜?｜는 절대값을 의미한다.In Equation 1, input nodes i, i = 1, 2, ..... N, N are input node numbers, output nodes j, j = 1, 2, .... M, M are output nodes It is a number. In addition, W _j and w _ij are N-dimensional weight vectors for the j th output node and weights between the i th input node and the j th output node, respectively, and X and x _i are the N th input vector and the i th input node, respectively. to be.

Is the weighted infinity norm, E = diag (1 / ε ₁ , 1 / ε ₂ , .... 1 / ε _N ), which is the diagonal weight matrix in the N × N dimension, and ε _i is the i-th input node Boundary argument for the input pattern of 이며? |

Only nodes having the minimum distance calculated by Equation 1 among the output nodes are activated, and the similarity between the input pattern and the template pattern for the activated output node is determined by checking boundary argument conditions in layer 2. In the ART2 neural network having the boundary factor, the boundary factor condition shown in Equation 2 below is used.

경계인수 조건:

Boundary argument condition:

In Equation 2, W _J is an N-dimensional template pattern vector for the activated J th output node, and J is an activated output node.

If the input pattern passes the boundary factor check, it is learned in the same class using the weight adjustment equation as shown in Equation 3 below.

W _J ^new = {X + W _J ^old [class _J ^old ]} / {[class _J ^old + 1]}

In Equation 3, W _J ^old and W _J ^new are weights before and after adjustment at the J th output node, respectively, and [class i] is the number of patterns belonging to class i. However, if it does not pass the boundary argument check, it saves the input pattern in a new class.

In Equation 1 and Equation 2, ε _i are boundary arguments for determining the degree of correspondence between the input pattern and the template pattern of the active output node, and whether to classify the two patterns as the same class or classify them as a new class. It is a value that determines whether or not. Among the estimated parameters as input patterns, ε _i is set large for parameters that vary greatly due to failure, and small values for smaller parameters can be classified by considering the sensitivity difference of each parameter according to the failure type. For example, when N = 2 and x ₁ changes larger than x ₂ , ε ₁ is set larger than ε ₂ .

7A to 7D are diagrams illustrating an example of a failure diagnosis experiment and an automatic level inspection test of an induction motor using a press process failure diagnosis apparatus according to the present invention.

First, as shown in Figure 7a, for the failure diagnosis experiment of the induction motor, the low-pressure three-phase induction motor is composed of three electric motors of steady state, rotor failure, bearing failure, respectively. In addition, the data acquisition unit uses the NI9234, and equipped with a vibration sensor for vibration detection.

FIG. 7B illustrates the ultrasonic level sensor used for the automatic lubricant test, and FIG. 7C illustrates an example of a combination of an induction motor failure diagnosis and an automatic lubricant level test. FIG. 7D is a failure diagnosis in a GUI (Graphic User Interface) state. And an example showing an automatic lubricating oil level inspection system.

Hereinafter, a simulation result using the press process failure diagnosis apparatus according to the present invention using the system of FIGS. 7A to 7D will be described.

Here, two types of failures of induction motors can be cited. In the present invention, the following two types of failures are artificially generated to perform failure diagnosis experiments of induction motors.

First, fault # 1 (Fault # 1) is a rotor fault, which artificially drilled a hole in the center bar of the rotor with a drill to cause a failure and performed a simulation (FIG. 8A).

Second, failure # 2 is a bearing failure, in which a powder was put into the bearing to cause wear and simulation was performed (FIG. 8B).

Accordingly, as shown in Table 1, the input of the ART2 neural network fault classifier uses the 16 vibration signal data measured and frequency-converted by the induction motor, and sets the boundary factor? Of the fault classifier. ε = [0.03, 0.01, 0.01, 0.01, 0.01, 0.01, 0.05, 0.001, 0.01, 0.001, 0.02, 0.01, 0.01, 0.01, 0.01, 0.01}

9A to 9C are graphs showing spectrums of FFT-converted vibration signal data for steady state, fault # 1, and fault # 2, respectively, and are relatively at fault # 1 and fault # 2 than the normal state. It can be seen that the magnitude of the maximum amplitude is large.

FIG. 10A shows a fault diagnosis result when the induction motor is in a normal state (P01) and a fault # 1 (P11) occurs in the induction motor at the 100th moment. The fault # 1 pattern P11 is connected to the ART-2 neural network. If it is entered, it does not pass the boundary argument test with class # 1 (normal state), so it automatically generates new class # 2 and declares it as new fault # 1 (rotor fault). Next, it can be seen that the fault diagnosis result when the new fault # 2 (bearing fault) occurs in the induction motor at the 100th moment is shown in FIG. 10B and generates a new class # 3.

That is, when the steady state pattern P01 is input, it can be seen that these patterns pass the boundary factor check and are classified into the failure class # 1. Next, when fault # 1 pattern (P11) is input, these patterns pass the boundary factor test with class # 2 and are classified as fault class # 2 to diagnose the current state of induction electric machine as fault # 1 (rotor fault). It can be seen. Likewise, when the fault # 2 pattern (P21) is input, it can be seen that it is classified as the fault class # 3.

11A to 11C show the fault diagnosis test results for the motor causing the (a) steady state, (b) rotor failure, and (c) bearing failure in the fault diagnosis and automatic inspection GUI (Graphical User Interface) program, respectively. This shows that the fault diagnosis system is correctly diagnosed.

12A to 12C are diagrams showing the results of lubricating oil level inspection in the fault diagnosis and automatic inspection GUI programs, respectively, (a) showing that the measured lubricating oil level is higher than the set level (reference value) and is displayed as normal. In b), it can be seen that the warning light is turned on because the measured lubricant level is lower than the set level. 12C shows that the state of the motor and the level of the lubricating oil are normal.

As a result of the simulation using experimental data in the actual induction motor and lubricant level test, it can be seen that the present invention is well applied to the press process diagnosis.

Therefore, if the present invention is applied and operated in the actual press process system, it is expected to greatly contribute to improving the reliability of the system, and can be applied to the failure diagnosis process in a multi-fault situation.

In the above, the failure diagnosis of the induction motor has been described, but it is applicable to the failure diagnosis of the press process system as well as various types of motors other than the induction motor, and the present invention is not limited to the above embodiments, Therefore, even if a person skilled in the art does design modification or avoidance design within the range without departing from the scope of the technical idea of the present invention will be within the scope of the present invention.

1 is a functional block diagram showing the configuration of a pressing process failure diagnosis apparatus according to a preferred embodiment of the present invention,

2 is a flowchart showing a press process failure diagnosis process through the press process failure diagnosis apparatus of FIG. 1;

3A and 3B are graphs illustrating a frequency change process of the frequency analyzer, and a graph illustrating a max extraction process and a measurement data FFT conversion process in units of 100 Hz, respectively.

4 is a view showing a frequency setting portion to be used as an actual input to the ART2 neural network according to the present invention,

5 is a structural diagram of a fault classifier using the ART2 neural network according to the present invention;

6 is a flowchart summarizing the operation of the fault classifier using the ART2 neural network in the fault diagnosis apparatus according to the present invention;

7a to 7d is a view showing an example of the press process failure diagnosis experiment using the press process failure diagnosis apparatus according to the present invention,

8 and 8bs are diagrams showing two types of failures artificially generated in the present invention, each showing a rotor failure and a bearing failure.

9A to 9C are graphs showing spectrums of FFT-converted vibration signal data according to the present invention;

10A and 10B are graphs showing fault diagnosis results at fault # 1 and fault # 2, respectively;

11a to 11c is a view showing a failure diagnosis and lubricant level automatic inspection GUI (failure diagnosis experiment) of the induction motor according to the present invention,

12A to 12C are diagrams showing a failure diagnosis and lubricant level automatic inspection GUI (lubricant level experiment) of an induction motor according to the present invention;

* Description of Major Symbols in Drawings *

110: induction motor

120: vibration sensor

130: data collector

140: frequency analysis unit

150: fault sorter

Claims (15)

Vibration sensor for detecting the vibration signal from the motor, A lubricant level sensor for measuring the lubricant level in the lubricant tank, A data collector configured to digitally store and store the vibration signal detected by the vibration sensor; A frequency analyzer for converting the vibration signal data stored in the data collector into a frequency domain; The vibration signal data converted by the frequency analyzer is input as an input pattern and does not pass the boundary argument test with class 1 to generate class 2, and then another pattern is input and does not pass the boundary argument test with class 2 A classifier for generating a class 3, wherein the class 1 is a normal state, the class 2 is a rotor failure, and the class 3 is a failure classifier. And a lubricating oil level automatic inspection unit for determining whether the level of the lubricating oil measured by the lubricating oil level sensor is broken in an on / off manner compared to a reference value. The method of claim 1, The lubricating oil level sensor is a failure diagnosis device, characterized in that the ultrasonic level sensor of the sensor unit and the transmitter display unit separated form. The method of claim 1, And the frequency analyzer converts the vibration signal data into a frequency domain. The method of claim 3, The frequency analyzer divides the FFT-converted data into 100 Hz frequency units, and extracts one signal having a maximum amplitude value from the divided 100 Hz units for a monitoring frequency band. . (a) detecting a vibration signal in the motor; (b) digitalizing and storing the vibration signal; (c) converting the vibration signal data into a frequency domain; (d) The vibration signal data converted into the frequency domain is inputted as an input pattern and does not pass the boundary argument test with class 1 to generate class 2, and then another pattern is input and the boundary argument test with class 2 is performed. Class 3 fails to pass, class 1 is a normal state, class 2 is a rotor failure, and class 3 a bearing failure, (e) a method for diagnosing a failure, characterized in that the automatic check of the lubricant level determines whether the failure level in the on / off manner compared to the reference value for the level of the lubricant measured by the lubricant level sensor. The method of claim 5, In the step (c), FFT converts the vibration signal data into a frequency domain. The method of claim 6, In the step (c), And dividing the FFT-converted data into each 100 Hz frequency unit, and performing a method of extracting one signal having a maximum amplitude value from the separated 100 Hz unit for a monitoring frequency band. Diagnostic method. delete delete delete delete delete delete delete delete

Images