KR101280174B1 - Method of detecting defect of an object using a high frequency signal - Google Patents

Abstract

The present invention relates to a method for determining a defect of an object, for example a rotating blade, by using a high frequency signal of the collected signal. The defect determining method may include separating a collection signal having defect characteristic information of the object into mono-component signals, selecting a high frequency signal among the mono-component signals as a characteristic signal, and analyzing the selected characteristic signal. Building a defect determination model.

Rotating Blades, Defects, Hidden Markov Models, Hilbert Sulfur Transformation

Description

Object defect determination method using high frequency signal {METHOD OF DETECTING DEFECT OF AN OBJECT USING A HIGH FREQUENCY SIGNAL}

The present invention relates to a method for determining a defect of an object, for example a rotating blade, by using a high frequency signal of the collected signal.

The presence of defects such as cracks in rotating blades, such as turbine blades, rotorcraft, etc., can cause problems in the operation of the rotating structure itself, as well as affect the overall installation and production.

Therefore, a method for determining a defect of the rotating blade was required, and as a result, a defect determination model appeared.

However, the conventional defect determination model was able to accurately reflect the position or depth of the defects that were clearly revealed, but did not accurately reflect the position or depth of the defects. As a result, it was difficult for the user to confirm the presence of the microscopic defects and the effects thereof.

In addition, since the defect determination model is complicated, it is difficult for a user to intuitively grasp the position of the defect and the like through the defect determination model. That is, it is inconvenient for the user to use the defect determination model.

These problems have occurred not only in the rotating blade but also in determining a defect of a general object.

It is an object of the present invention to provide a defect determination method that accurately reflects and intuitively predicts a defect even when a minute defect exists in an object, for example, a rotating blade.

In order to achieve the above object, a defect determination method of an object according to an embodiment of the present invention comprises the steps of: separating the collected signal having the defect characteristic information of the object into mono-component signals; Selecting a high frequency signal among the mono-component signals as a characteristic signal; And constructing a defect determination model using the selected high frequency signal.

According to another aspect of the present invention, there is provided a method of determining a defect of an object, including: acquiring a collection signal having defect characteristic information of the object; Selecting a peak value of the collected signal as a characteristic value; And determining a defect of the object by using the selected characteristic value. Here, the peak value is a peak value of a high frequency signal among the collected signals.

The defect determination model of the present invention according to the present invention accurately distinguishes the presence or absence of a defect, the position of the defect, the depth of the defect, and the like, and thus accurately determines the state of an object, for example, a rotating blade, which reacts sensitively to minute defects. can do.

In addition, since the user can accurately and intuitively determine the presence or absence of a defect, the location of the defect, the depth of the defect, and the like, the user's convenience can be improved and the device using the object can be secured.

In addition, using a hidden Markov model as a defect determination model, the defect determination model can be both functional and efficient, since only learning about a new pattern can be added when a non-trained pattern is found, thereby constructing a new hidden Markov model. .

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Like reference numerals are used for like elements in describing each drawing.

When a component is referred to as being "connected" or "connected" to another component, it may be directly connected to or connected to that other component, but it may be understood that other components may be present in between. Should be. On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the contextual meaning of the related art and are to be interpreted as either ideal or overly formal in the sense of the present application Do not.

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

1 is a view showing a rotary blade system used in the defect determination method according to an embodiment of the present invention.

The defect determination method of the present embodiment means a method of detecting a defect of a specific object, for example, a rotating blade, for example, a crack, and the like, and a peak value of a high frequency signal or a specific signal as described below. I use it. Here, the object is not limited to a specific object and includes all objects in which a defect can be generated. However, for convenience of description below, the object is assumed to be a rotary blade sensitive to defects.

Referring to FIG. 1, an object defect determining system is a system used in a method of determining a defect, for example, a crack, of an object 102, for example, a rotation blade, and includes a rotating shaft 100 and a rotating blade ( 102, a sensor 104, and a defect determination device 106.

The rotary blade 102 rotates according to the rotation of the rotary shaft 100 and includes all the rotating blades such as a propeller of a helicopter and a turbine blade.

Sensor 104 is installed at the end of the rotary blade 102, for example, as shown in Figure 1 serves to detect the cracks present in the rotary blade 102. According to one embodiment of the invention, the sensor 104 is an acceleration sensor or a displacement sensor.

The sensor 104 is connected to the defect determining device 106 by wire or wirelessly, and transmits a signal having a detection result, for example, a vibration signal having vibration information of the rotating blade 102, to the defect determining device 106. do.

In this case, the vibration signal is used as a defect determination method because the vibration characteristics of the rotary blade 102 vary depending on the position and depth of the defect when the defect is present or not in the rotary blade 102. . According to an embodiment of the present invention, the system manager detects a vibration signal while rotating the rotary blade 102 or when the impact is applied to the rotary blade 102 while being fixed without rotating the rotary blade 102. Detect vibration signal.

In the above, the sensor 104 has been described as transmitting the vibration signal indicating the vibration information of the rotary blade 102 to the defect determining device 106, but is not limited to the vibration signal. However, for convenience of description, the signal collected and transmitted by the sensor 104 will be referred to as a collection signal.

The defect determination device 106 receives many collected signals transmitted from the sensor 104 and analyzes the received collected signals to build a defect determination model. Therefore, if the defect determination model thus constructed is installed in a helicopter or the like, it is possible to detect the presence or absence of the defect and the location of the defect in real time.

Hereinafter, a process of building the defect determination model will be described in detail with reference to the accompanying drawings. However, the defect is assumed to be a crack for convenience of explanation.

2 is a flowchart illustrating a defect determination model construction and a defect determination process according to an embodiment of the present invention.

2, the system of the structure shown in Figure 1 for the crack detection of the rotary blade 102 is installed (S200).

Subsequently, the sensor 104 transmits a collection signal including vibration information according to the rotation or impact of the rotary blade 102 to the defect determination device 106 (S202). Specifically, the model builder experiments separately when there is no crack in the rotary blade 102 and when there is a crack in the rotary blade 102. The model builder also experiments with varying the location and depth of the cracks. As a result, collected signals in various environments are transmitted from the sensor 104 to the defect determination device 106.

Subsequently, the defect determination apparatus 106 preprocesses the collected collection signals (S204). Specifically, the defect determination device 106 extracts the send feature vectors from the collected signals and signs the feature vectors. For example, the defect determination device 106 extracts the peak values of each collected signal as a feature vector, quantizes the extracted feature vector, and signs it.

According to an embodiment of the present invention, the defect determination method uses an Empirical Mode Decomposition (EMD) method of the Hilbert Huang Transform to extract feature vectors during the preprocessing. A detailed description thereof will be described later.

Subsequently, the defect determination device 106 selects a part of the collected signal as the training data, and constructs a probability relationship using the selected data using a specific learning algorithm (S206).

According to an embodiment of the present invention, the Baum-Welch algorithm is used as the learning algorithm, and the defect determination method generates a statistical structure of a hidden Markov model through the learning.

Subsequently, the defect determination apparatus 106 constructs a defect determination model, for example, a hidden Markov model, using the statistical structure collected through the learning process (S208).

Subsequently, the defect determination method determines a defect of the rotating blade 102 using the constructed defect determination model (S210). Specifically, the defect determination method accurately determines the presence or absence of the defect, the position and the depth of the defect.

In short, the defect determination method of the present invention collects collection signals in various environments and preprocesses and learns the collected collection signals to build a defect determination model.

Hereinafter, detailed steps will be described in detail with reference to the accompanying drawings.

1. Pretreatment process (S204)

3 is a flowchart illustrating a pretreatment process according to an embodiment of the present invention, and FIGS. 4 to 6 are diagrams showing experimental results of a pretreatment process in various environments.

Referring to FIG. 3, the defect determination method separates an acquisition signal into a plurality of mono-component signals using an empirical mode separation method (EMD) of Hilbert sulfur conversion (S300). For example, the collected signal as shown in Fig. 4A is separated into mono-component signals as shown in Fig. 4B.

In the actual experiment, the experiment is conducted under various conditions in order to detect the presence, position, and depth of the crack of the rotary blade 102. For example, as shown in FIG. 5 (A), after setting specific conditions equally, for example, the rotation speed at which the cracks are formed at the same position of the rotating blade 102 may be different, or in FIG. 5 (B). Experiment with varying the position of the crack at the same rotational speed as shown. As a result, the defect determination device 106 can collect the collection signals in various environmental conditions. The defect determination device 106 then separates each of the collected signals into mono-component signals using, for example, heuristic mode separation.

However, as shown in FIG. 5, since some of the collected signals have a similar signal pattern, it is difficult to accurately and intuitively determine the existence and location of the crack by using the collected signals themselves. Therefore, the defect determination method of the present invention uses a high frequency signal among the mono-component signals of the collected signal as described below.

Referring again to Figs. 4B and 4C, the collected signal is separated into mono-component signals of c1 to c5. Here, it can be seen that the collected signals have non-linear characteristics, whereas the mono-component signals have linear characteristics.

Looking at the characteristics of the mono-component signals, c1 of the mono-component signals are smaller in size than c2 to c5 but contain relatively high frequency components, and c2 to c5 are larger in size than c1 but are relatively low frequency. Contains ingredients. Here, when synthesizing c2 to c5, it can be seen that a pattern similar to the collected signal appears as shown in FIG. 4 (C). That is, it can be seen that c1 of c1 to c5 represents unique characteristics of the collected signal.

Therefore, the defect determination method of the present invention selects a high frequency signal among the mono-component signals separated from the collected signal as a characteristic signal (feature vector) of the collected signal (S302).

Subsequently, the mono-component signal selected as the characteristic signal is Fourier transformed (S304). For example, if the c1 selected in FIG. 5B is Fourier transformed, a frequency signal IMF # 1 as shown in FIG. 6A can be obtained.

Referring to FIG. 6 (A), it can be seen that the Fourier transform of the original collected signal does not produce a distinctive feature, whereas the Fourier transform of the characteristic signal c1, which is a high frequency signal, clearly distinguishes and outputs peak values. For example, distinct peak values were obtained at 0.5 Hz, 10 Hz, 20 Hz and 40 Hz.

As a result of the experiment, the peak values were clearly distinguished from each other in various environments, such as when there is no crack, when cracks are present, when cracks are different, and when cracks are different. Therefore, the defect determination method of the present invention can determine the presence of the crack, the position of the crack and the depth of the crack as described later through the comparison of these peak values.

Subsequently, the defect determination method normalizes the Fourier transformed mono-component signal on the basis of the maximum peak value (S306). For example, the mono-component signal is normalized to have a maximum magnitude of 1, so that the influence of the frequency component due to the speed change of the rotating blade 102 is not taken into account.

In the above, the preprocessing process for one acquisition signal has been described, but in the actual defect determination process, the pretreatment process for the acquisition signals collected in various environments is performed.

2. Learning and defect determination model building process ( S208 )

7 is a diagram illustrating a learning process according to an embodiment of the present invention, and FIG. 8 is a diagram illustrating a defect determination model building process according to an embodiment of the present invention.

The learning process means, for example, a process of constructing a probabilistic relationship between each state and an observation symbol in a hidden Markov model.

Specifically, as shown in FIG. 7, some of the collected signals are selected as learning data, and a probability relationship is configured using, for example, a Baum-Welch algorithm.

Based on the results obtained in the above learning, we construct a hidden Markov model as a defect judgment model, for example.

For example, a hidden Markov model is constructed with state 800 and observation symbol 802 as shown in FIG. Specifically, state 800 may be divided into a case where there is a crack and a case where there is no crack, and the observation symbol 802 indicates collected signals obtained in various environments. Here, the probability that a crack does not exist and the probability that a crack exists in each observation symbol 802 are matched, respectively. For example, there is a 5% probability that a crack exists in the observation symbol "A" and a 95% probability that a crack does not exist.

3. Defect Determination Process (S210)

Using the defect determination model constructed as described above, for example, the hidden Markov model, it is possible to determine the presence or absence of the crack, the position of the crack, and the depth of the crack. Here, the determination result may be displayed as a probability. According to an embodiment of the present invention, the determination result may be represented by a log-likelihood.

Hereinafter, various defect determination processes will be described.

9 is a diagram illustrating a result of determining whether a defect exists using a defect determination model according to an embodiment of the present invention. Specifically, FIG. 9 illustrates a result of experiments by applying a hidden Markov model for determining the presence of cracks on the rotating blade 102 and applying the same. Here, the determination result is expressed in log wood, which means the probability that the observed symbol permutation will exist in the hidden Markov model.

9 (A) shows the results of normalizing the extraction of the feature vector while sequentially changing the speed of the rotating blade 102 from 20 rad / s to 90 rad / s when there is a crack in the rotating blade 102 and the rotating blade 102. In the absence of cracks, the feature vector is extracted and normalized while sequentially changing the speed of the rotating blade 102 from 20 rad / s to 80 rad / s.

The result of learning this feature vector is shown in Fig. 9B.

Looking at the learning results in the absence of cracks, in spite of the various rotational speeds of the rotating blades 102, the judgment result was that a log wood close to zero was obtained for the observed symbol permutation without the cracks. 102 indicates that there is a high probability that no crack exists. That is, when there is no crack in the rotary blade 102, since the probability of the signal without the crack shows close to 0, it can be confirmed that it is accurately determined that there is no crack.

Looking at the learning result in the presence of cracks, despite the various rotational speeds of the rotary blade 102, the determination result is that the logwood is close to -100, which indicates that there is a high probability that the crack exists in the rotary blade 102. it means. That is, when the crack is present on the rotary blade 102 shows that the probability of the signal without the crack is close to -100, it can be confirmed that it is accurately determined that the crack exists.

In other words, since the log wood when there is a crack and when there is no crack is clearly distinguished, the user can determine whether the crack exists accurately and intuitively using the defect determination model.

10 is a view showing a result of a defect position determination using a defect determination model according to an embodiment of the present invention.

As shown in FIG. 10 (A), the rotary blade 102 (set the total length to 1) was separated into a fixed end and a free end based on 0.4 to determine a defect position. Herein, when the crack exists in the range of 0 to 0.4 in the rotary blade 102, it is represented by the crack of the fixed end, and when the crack exists in the range of 0.4 or more, it is represented by the crack of the free end.

As a result of constructing a hidden Markov model based on the collected signals when there is a crack at a fixed stage of 0.4 or less, the determination result was obtained as a log wood close to zero as shown in FIG. 10 (C).

On the other hand, if there is a crack in the free end of 0.5 or more, the hidden Markov model is constructed based on the collected signals.

That is, it can be confirmed that the crack accurately reflects whether the crack is located at the free end or the fixed end according to the position of the crack.

FIG. 11 is a diagram illustrating a defect depth determination result using a defect determination model according to an embodiment of the present invention.

As shown in FIG. 11 (A), the crack depth was determined after separating the crack (whole set to 1) into 0.1 or less, 0.1 to 0.3, 0.3 or more.

Feature vectors were obtained as shown in FIG. 11 (B), and the determination result was obtained as shown in FIG. 11 (C).

As shown in FIG. 11C, the log wood appeared differently depending on the depth of the crack. Specifically, when the crack depth is 0.1 or less, the judgment result has a log wood close to zero, and when the crack depth is 0.1 to 0.3, the judgment result has a log wood close to about -10. If the depth is 0.3 or more, the judgment result is a logwood close to -100.

In other words, it can be seen that the hidden Markov model accurately reflects the crack depth.

In short, the defect determination method of the present invention constructs a defect determination model using the peak values of the collected signals, and displays the determination result as a probability through the constructed defect determination model.

However, the presence or absence of a crack, a position determination, and a depth determination method may be variously performed.

According to an embodiment of the present invention, after establishing a defect determination model for determining the presence of cracks to determine the presence of cracks, if a crack is found after building a defect determination model for determining the position of the cracks If the position of the crack is determined, and then the position of the crack is determined, a defect determination model for determining the depth of the crack may be constructed and then the depth of the crack may be determined.

According to another embodiment of the present invention, the feature may be determined after constructing a defect determination model capable of determining at least two of presence or absence of cracks, position of cracks, and depth of cracks at once.

Hereinafter, the process of judging two or more features at a time will be described.

12 is a diagram illustrating a process of determining whether a crack is present and a position of a crack at the same time according to an embodiment of the present invention.

Referring to FIG. 12 (B), if there is no crack, the log wood is close to -100, whereas if there is a crack, the log wood has a log wood of -10 or less. Specifically, log wood of -4 or less is detected when the crack is located at a fixed end of 0.4 or less while cracks are present, and -4 to -10 when cracks are located on the free end of 0.4 or more while the crack is present. Logwood in between was detected.

In other words, it can be seen that the hidden Markov model accurately reflects the existence and location of cracks at once.

Although not shown above, the defect determination model of the present invention may be constructed by accurately reflecting the existence of cracks and the depth of cracks, or may be constructed by accurately reflecting the presence of cracks, the position of cracks, and the depth of cracks. .

In short, the defect determination method of the present invention extracts a high frequency signal from a collected signal, Fourier transforms the extracted high frequency signal, and builds a defect determination model by learning the peak values of the converted high frequency signal. Then, the defect determination model is used to determine the defect of the rotary blade 102, in particular, the presence or absence of a crack, the position of the crack, and the depth of the crack.

As described above, when using the defect determination model of the present invention, in particular the hidden Markov model, the presence or absence of defects, the position of the defects and the depth of the defects are accurately distinguished, so that the rotating blade that rotates sensitively to fine defects ( The state of 102 can be accurately determined.

In addition, since the user can accurately and intuitively determine whether there is a defect, the location of the defect, the depth of the defect, and the like through the defect determination model, the user's convenience is improved and the device of the rotating blade 102 is used. I can aim for safety.

In addition, using the hidden Markov model as a defect determination model may only be functional and efficient since adding a learning about a new pattern only results in a new hidden Markov model when an untrained pattern is found.

In the above, the rotary blade 102 as an example of the defect determination, but the defect determination method of the present invention can be applied to any object that can determine the defect using a high frequency signal or a peak value.

The embodiments of the present invention described above are disclosed for purposes of illustration, and those skilled in the art having ordinary knowledge of the present invention may make various modifications, changes, and additions within the spirit and scope of the present invention. Should be considered to be within the scope of the following claims.

1 is a view showing a rotating blade system used in the defect detection method according to an embodiment of the present invention.

2 is a flowchart illustrating a defect determination model construction and a defect determination process according to an embodiment of the present invention.

3 is a flowchart illustrating a pretreatment process according to an embodiment of the present invention.

4 to 6 are diagrams showing experimental results of a pretreatment process in various environments.

7 is a diagram illustrating a learning process according to an embodiment of the present invention.

8 is a diagram illustrating a defect determination model building process according to an embodiment of the present invention.

9 is a diagram illustrating a result of determining whether a defect exists using a defect determination model according to an embodiment of the present invention.

10 is a view showing a result of a defect position determination using a defect determination model according to an embodiment of the present invention.

FIG. 11 is a diagram illustrating a defect depth determination result using a defect determination model according to an embodiment of the present invention.

12 is a diagram illustrating a process of determining whether a crack is present and a position of a crack at the same time according to an embodiment of the present invention.

Claims (12)

In the defect determination method of the object, Separating the collected signal having defect characteristic information of the object into mono-component signals; Selecting a high frequency signal among the mono-component signals as a characteristic signal; And And constructing a defect determination model using the selected high frequency signal. The method of claim 1, wherein the object is a rotating blade, the collection signal is a vibration signal, and the mono-component signals are linear signals, The vibration signal is a defect determination method of the object, characterized in that the signal sensed by the sensor during the rotation of the object or the impact on the object. The method of claim 1, wherein the building of the defect determination model comprises: Fourier transforming the characteristic signal; And Normalizing the characteristic signal based on the maximum peak value of the Fourier transformed characteristic signal, And the defect of the object is detected by comparing peak values of the Fourier transformed characteristic signal. The method of claim 1, wherein the separating into mono-component signals comprises: Separating the first collected signal into first mono-component signals indicative of the presence of cracks in the object; Separating the second collected signal indicative of the location of the crack present in the object into second mono-component signals; And Separating the third acquisition signal indicative of the depth of crack present in the object into third mono-component signals, The selecting of the characteristic signal may include extracting characteristic signals from the first collection signal, the second collection signal, and the third collection signal, respectively. The method of claim 1, wherein the separating into mono-component signals comprises: Separating the collected signal into the mono-component signals indicative of the presence or absence of a crack in the object, the location of the crack and the depth of the crack into the mono-component signals. Judgment method. 2. The method of claim 1, wherein the empirical mode separation method of Hilbert sulfur conversion is used for separation into the mono-component signals, and a hidden Markov model is used for constructing the defect determination model. The method of claim 1, wherein a plurality of acquisition signals indicative of defect characteristics of the object are obtained, The defect determination method, Selecting some of the collected signals as learning data; And Training the selected collected signals using a specific learning algorithm; And The method may further include constructing the defect determination model based on the learned data. The defect determination model indicates the presence or absence of cracks, the position of the cracks and the depth of the cracks as a probability, the probability is represented by a log wood. In the defect determination method of the object, Obtaining a collection signal having defect characteristic information of the object; Selecting a peak value of the collected signal as a characteristic value; And Determining a defect of the object by using the selected characteristic value; And the peak value is a peak value of a high frequency signal among the collected signals. The method of claim 8, wherein the object is a rotating blade and the collection signal is a vibration signal detected by a sensor when the object is rotated or when the object is impacted. The method of claim 8, wherein the selecting of the peak value as a characteristic value uses a Hilbert sulfur conversion method, and the defect of the object is determined through a defect determination model constructed based on the characteristic value. And the defect determination model uses a hidden Markov model. The method of claim 8, wherein a plurality of acquisition signals are obtained, The collection signals may include a first collection signal indicating whether a crack exists in the object, a second collection signal indicating a location of the crack, and a third collection signal indicating a depth of the crack. The defect determination method of the object made into. The method of claim 8, wherein the collection signal displays information on at least two of presence or absence of a crack, a position of the crack, and a depth of the crack.

Images