JP2021162458A - Tank inspection device - Google Patents

Abstract

To provide a tank inspection device capable of detecting a state of a reinforcing layer of a high-pressure tank.SOLUTION: A tank inspection device 20 includes a computing device that has an extracting portion 121 that extracts a plurality of AE waveforms from output waveforms of an AE sensor 3 detecting AE waves generated in a high-pressure tank 50, a sorter 123 that mechanically learns so as to classify the plurality of AE waveforms extracted by the extracting portion 121 into first waveforms derived from transverse cracks, second waveforms derived from micro-cracks, third waveforms derived from an interlayer peeling, and fourth waveforms derived from fiber breakage, a counting portion 124 that counts each number of the first waveforms to the fourth waveforms, and an output portion 125 that outputs results counted by the counting portion 124.SELECTED DRAWING: Figure 2

Description

The present invention relates to a tank inspection apparatus for inspecting a high-pressure tank including a liner and a reinforcing layer made of a fiber reinforced resin covering the outer surface of the liner.

Conventionally, as an inspection method for a tank containing a liquid, an inspection method using an AE (acoustic emission) method is known (see, for example, Patent Document 1). Patent Document 1 describes a test device for a composite material tank made of a liner made of aluminum or the like and a fiber reinforced material covering the outer surface of the liner, and a test device for inspecting the damaged state of the liner using an AE sensor. ing. In this test apparatus, the AE signal is received by the AE sensor, and the damaged state of only the liner is detected from the frequency and energy change of the AE signal.

Japanese Unexamined Patent Publication No. 2015-031630

By the way, a high-pressure tank provided with a liner and a reinforcing layer made of a fiber-reinforced resin covering the outer surface of the liner is filled with a high-pressure (for example, 70 MPa) gas, so that very high strength is required. When designing such a high-pressure tank, various manufacturing conditions (wrapping of the fiber-reinforced resin layer) when the material constituting the reinforcing layer and the fiber-reinforced resin layer composed of the fiber bundle impregnated with the resin are wound around the outer surface of the liner. A high-pressure tank is manufactured by changing the angle and curing conditions). Then, it is necessary to pressurize the inside of each high-pressure tank to a predetermined pressure, cut the high-pressure tank at a plurality of places, and observe the cross section to confirm the state of the high-pressure tank.

However, in this method, since it is necessary to cut the high-pressure tank and observe the cross section in order to confirm the state of the high-pressure tank, the confirmation work takes a very long time. Further, it is difficult to accurately confirm the state of the high-pressure tank because, for example, the state of occurrence of cracks and delamination differs depending on the cutting position. Therefore, there is a problem that it takes a very long time to design a high-pressure tank.

In the test apparatus described in Patent Document 1, since the damaged state of only the liner is detected, the state of the reinforcing layer cannot be detected, and the type of damage cannot be detected.

The present invention has been made in view of these points, and an object of the present invention is to provide a tank inspection device capable of detecting the state of a reinforcing layer of a high-pressure tank.

The tank inspection device according to the present invention is a tank inspection device that inspects a high-pressure tank including a liner and a reinforcing layer made of a fiber-reinforced resin that covers the outer surface of the liner, and the tank inspection device is the inside of the high-pressure tank. A plurality of AE waveforms are extracted from the output waveform of the AE sensor that detects the AE wave generated in the high pressure tank when the pressure is increased, and the high pressure tank is inspected based on the extracted plurality of AE waveforms. The reinforcing layer is composed of a plurality of fiber-reinforced resin layers, and the tank inspection device extracts the plurality of AE waveforms from the output waveform of the AE sensor that detects the AE wave generated in the high-pressure tank. The first waveform derived from the transverse crack that passes through the fiber-reinforced resin layer in the thickness direction and the plurality of AE waveforms extracted by the extraction unit do not pass through the fiber-reinforced resin layer in the thickness direction. A classifier machine-learned to classify into a second waveform derived from microcracks, a third waveform derived from delamination of the fiber-reinforced resin layer, and a fourth waveform derived from fiber breakage, and the above. A computing device including a counting unit that counts each number of the first waveform to the fourth waveform classified by the classifier and an output unit that outputs the result of counting by the counting unit is included.

In the present specification and claims, AE is an abbreviation for acoustic emission, and refers to a phenomenon in which energy is released as elastic waves when a material or structure is broken or deformed. Further, the AE wave is an elastic wave emitted when a material or a structure is broken or deformed. Further, the AE sensor refers to a sensor that detects an elastic wave (AE wave) generated by the generation of AE.

According to the tank inspection apparatus of the present invention, the plurality of AE waveforms are extracted from the output waveform of the AE sensor, and the plurality of AE waveforms are obtained as a first waveform derived from a transverse crack and a first waveform derived from a microcrack. It is classified into two waveforms, a third waveform derived from delamination of the fiber-reinforced resin layer, and a fourth waveform derived from fiber breakage, and the number of each of the classified first waveform to fourth waveform is counted. .. As a result, the damage mode (transverse crack, microcrack, delamination, fiber breakage) of the high-pressure tank and the number of occurrences thereof can be confirmed based on the counted result, so that the state of the reinforcing layer can be confirmed. There is no need to cut the high pressure tank and observe the cross section. Therefore, the time required for checking the state of the reinforcing layer of the high-pressure tank can be shortened. Further, unlike the case of observing the cross section, the state of occurrence of cracks, delamination and the like does not differ depending on the cutting position, so that the state of the reinforcing layer can be confirmed accurately. As a result, the time required for designing the high-pressure tank can be significantly reduced.

According to the present invention, it is possible to provide a tank inspection device capable of detecting the state of the reinforcing layer of a high-pressure tank.

It is the schematic which shows the whole structure of the inspection system provided with the tank inspection apparatus which concerns on 1st Embodiment of this invention. It is a block diagram which shows the structure of the control part of the tank inspection apparatus which concerns on 1st Embodiment of this invention. It is sectional drawing which shows the structure of the high pressure tank of FIG. It is a figure for demonstrating the transverse crack, microcrack, delamination and fiber breakage which occur in a reinforcing layer. It is a figure which shows the output waveform of the AE sensor at the time of filling a liquid and pressurizing a high pressure tank. It is a figure which shows the teacher data used when the classifier of the tank inspection apparatus which concerns on 1st Embodiment of this invention is machine-learned. It is a figure which shows the flow of the pressure inspection method which concerns on 1st Embodiment of this invention. It is a figure which shows the output waveform of the AE sensor when gas is filled and the high pressure tank is pressurized. It is a block diagram which shows the structure of the control part of the tank inspection apparatus which concerns on 2nd Embodiment of this invention. It is a figure which shows the flow of the pressure inspection method which concerns on 2nd Embodiment of this invention.

(First Embodiment)
Hereinafter, the inspection system 1 provided with the tank inspection device 20 according to the first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a schematic view showing an overall configuration of an inspection system 1 provided with a tank inspection device 20 according to the first embodiment of the present invention.

As shown in FIG. 1, the inspection system 1 is a system for inspecting the high-pressure tank 50, and includes a pump 2 that pressurizes the inside of the high-pressure tank 50 and one or more AE sensors 3 attached to the outer surface of the high-pressure tank 50. A pressure sensor 4 for detecting the internal pressure of the high-pressure tank 50 and a tank inspection device 20 for controlling the drive of the pump 2 and detecting the state of the high-pressure tank 50 are provided. The inspection system 1 is arranged in an inspection room composed of a thick concrete wall or the like.

The pump 2 fills the inside of the high pressure tank 50 with a fluid (liquid or gas) to pressurize the inside of the high pressure tank 50. In this embodiment, the pump 2 supplies water to the inside of the high pressure tank 50. The pump 2 draws water from a water storage tank (not shown) and supplies it to the high-pressure tank 50. The liquid supplied to the high pressure tank 50 is not limited to water, and a liquid other than water may be used.

Further, a supply pipe 6 made of metal or the like through which water supplied to the high pressure tank 50 passes is connected to the pump 2. The supply pipe 6 is provided with a valve 7 capable of opening and closing the flow path. Further, a discharge pipe 8 made of metal or the like for discharging the water inside the high pressure tank 50 to the outside of the inspection room is connected to the high pressure tank 50. The discharge pipe 8 is provided with a valve 9 capable of opening and closing the flow path.

The AE sensor 3 detects an AE wave generated due to the occurrence of cracks in the reinforcing layer 52 described later in the high-pressure tank 50, and outputs the detection result as an output waveform to the tank inspection device 20. The AE sensor 3 is not particularly limited as long as it can detect the AE wave generated in the high pressure tank 50, but for example, a piezoelectric sensor or the like can be used.

Further, the AE sensor 3 is fixed at a predetermined position on the outer surface of the reinforcing layer 52. The number and fixing positions of the AE sensors 3 are not particularly limited, but here, they are fixed at a total of three locations, both ends and the center of the high-pressure tank 50 in the axial direction (longitudinal direction). As a result, no matter where the AE wave is generated in the high pressure tank 50, the AE wave can be detected by the AE sensor 3. Even when one AE wave is detected by two or more AE sensors 3, it can be determined from the waveform that it is the same AE wave by using a known technique, so that the number of AEs generated can be increased. It will not be counted twice.

In the present embodiment, since the inside of the high-pressure tank 50 is filled with liquid (water in this case), the AE wave (vibration) generated when AE occurs in the high-pressure tank 50 causes both the reinforcing layer 52 and water. It propagates and is detected by the AE sensor 3. At this time, the AE wave is reflected at the interface between the liquid and the liner 51 when propagating in the liquid, or the wave propagating in the reinforcing layer 52 and the wave propagating in the liquid are superimposed. Therefore, it is usually difficult to accurately classify a plurality of AE waveforms extracted from the output waveform of the AE sensor 3 into a first waveform and a second waveform, which will be described later. However, in the present embodiment, as will be described later, by using the machine-learned classifier 123, it is possible to accurately classify a plurality of AE waveforms into a first waveform and a second waveform.

The pressure sensor 4 is configured to be able to detect the internal pressure of the high pressure tank 50. The pressure sensor 4 may be provided inside the high-pressure tank 50, or may be provided in the supply pipe 6 connecting the pump 2 and the high-pressure tank 50. The pressure sensor 4 outputs the detection result to the tank inspection device 20.

The tank inspection device 20 is a device that inspects the state of the reinforcing layer 52 described later in the high pressure tank 50 when the inside of the high pressure tank 50 is increased to a predetermined pressure. As shown in FIG. 2, the tank inspection device 20 includes a control unit 21 for controlling the pump 2 and the like, a display unit 22 including a display panel and the like, and an operation unit 23 including buttons and the like operated by the operator. The display unit 22 includes, for example, the internal pressure of the high-pressure tank 50, the pressurizing time, the output waveform of the AE sensor 3, the operating status of the pump 2, the operating status of the valves 7 and 9, and the extraction unit 121 and the conversion unit 122, which will be described later. The output of the device 123, the counting unit 124, the output unit 125, and the like can be displayed.

Here, the tank inspection device 20 can control the drive and stop of the pump 2, and in particular, can stop the pump 2 based on the output of the AE sensor 3. Further, the tank inspection device 20 controls the opening / closing operation of the valves 7 and 9. The detailed structure of the tank inspection device 20 will be described later.

As shown in FIG. 3, the high-pressure tank 50 is, for example, a tank filled with high-pressure hydrogen gas mounted on a fuel cell vehicle. The gas that can be filled in the high pressure tank 50 is not limited to hydrogen gas. The high-pressure tank 50 is a substantially cylindrical high-pressure gas storage container having dome-shaped rounded ends. The high-pressure tank 50 includes a liner 51 having a gas barrier property and a reinforcing layer 52 made of a fiber reinforced resin that covers the outer surface of the liner 51.

The liner 51 is a resin or metal member that forms a storage space filled with high-pressure hydrogen gas. Openings are formed at both ends of the liner 51 in the longitudinal direction (axial direction), and a base 54 and an end boss 56 are provided, respectively. The base 54 and the end boss 56 are made by processing a metal material such as aluminum or an aluminum alloy into a predetermined shape. A supply pipe 6 and a discharge pipe 8 for filling and discharging hydrogen gas with respect to the accommodation space are connected to the base 54.

The reinforcing layer 52 covers the outer surface of the liner 51, and the fiber reinforced resin layer made of a fiber bundle impregnated with resin is wound around the liner 51 to cure the plurality of fiber reinforced resin layers. It is formed. The reinforcing layer 52 has a function of reinforcing the liner 51 to improve mechanical strength such as rigidity and pressure resistance of the high pressure tank 50. The reinforcing layer 52 is composed of a thermosetting resin and reinforcing fibers. As the thermosetting resin, it is preferable to use a thermosetting resin such as a phenol resin, a melamine resin, a urea resin, and an epoxy resin, and it is particularly preferable to use an epoxy resin from the viewpoint of mechanical strength and the like. As the reinforcing fiber, glass fiber, aramid fiber, boron fiber, carbon fiber and the like can be used, and it is particularly preferable to use carbon fiber from the viewpoint of light weight, mechanical strength and the like.

When the pressure inside the high-pressure tank 50 is increased to a predetermined pressure, the reinforcing layer 52 has a transverse crack that passes through the fiber-reinforced resin layer 52a of the reinforcing layer 52 in the thickness direction and a fiber that is smaller than the transverse crack. Microcracks that do not pass through the reinforced resin layer 52a in the thickness direction, delamination that occurs at the interface between the fiber reinforced resin layers 52a, and fiber breakage in which the fibers constituting the fiber reinforced resin layer 52a are broken occur (FIG. 4). reference). Along with this, AE waves corresponding to transverse cracks, microcracks, delamination and fiber breakage are generated in the reinforcing layer 52 of the high pressure tank 50.

Next, the detailed structure of the tank inspection device 20 will be described.

As shown in FIG. 2, the control unit 21 of the tank inspection device 20 includes a CPU and a storage unit including a ROM and a RAM. The CPU executes an operation program stored in the storage unit.

Further, the control unit 21 includes an extraction unit 121, a conversion unit 122, a classifier 123, a count unit 124, an output unit 125, a pressurization execution unit 126, and a valve opening / closing unit 127 as software. In the present embodiment, the "calculation unit" of the present invention is composed of the extraction unit 121, the conversion unit 122, the classifier 123, the count unit 124, and the output unit 125.

The output waveform of the AE sensor 3 is input to the extraction unit 121. The extraction unit 121 extracts a plurality of AE waveforms from the output waveform of the AE sensor 3. Specifically, the output waveform of the AE sensor 3 is, for example, a waveform obtained by connecting a large number of waveforms as shown in FIG. In the waveform shown in FIG. 5, the region A represents the vibration generated when the AE is generated in the high pressure tank 50. Further, the region A1 represents the vibration propagating in the reinforcing layer 52, and the region A2 represents the vibration propagating in the water.

When AE occurs in the high-pressure tank 50 and the amplitude becomes equal to or higher than a predetermined threshold value (L), the extraction unit 121 extracts a mass of vibrations (region A) within a certain period including the timing as one AE waveform. The threshold is set for noise removal. When a predetermined time elapses after the amplitude becomes equal to or higher than a predetermined threshold value, the vibration becomes smaller and the amplitude becomes less than the predetermined threshold value. After extracting the AE waveform, the extraction unit 121 extracts the next AE waveform when the amplitude becomes equal to or higher than a predetermined threshold value again after a certain period of time has elapsed. In this way, the extraction unit 121 sequentially extracts a plurality of AE waveforms from the output waveform of the AE sensor 3 and outputs them to the conversion unit 122.

When AE waveforms caused by a plurality of AEs are connected and output as one large (long time axis) AE waveform, even if the waveforms are classified as described later by the classifier 123, they are classified with high accuracy. Can not do it. Therefore, if the amplitude does not fall below the predetermined threshold even after a certain period of time has passed after the amplitude exceeds the predetermined threshold, the AE waveform is not classified by the classifier 123. Since such an AE waveform is less than a few percent of all the AE waveforms, it has almost no effect on the count number of the AE waveform, that is, the inspection result of the high pressure tank 50.

Each AE waveform extracted by the extraction unit 121 includes a first waveform derived from the occurrence of a transverse crack that increases immediately before the high-pressure tank 50 is destroyed, and a second waveform derived from the occurrence of a microcrack smaller than the transverse crack. , At least a third waveform derived from delamination and a fourth waveform derived from fiber breakage, but the AE waveform immediately after extraction is divided into the first waveform, the second waveform, the third waveform, and the fourth waveform. It is not easy to classify. Therefore, in the present embodiment, the extracted AE waveform is frequency-analyzed by the conversion unit 122, and then classified into the first waveform to the fourth waveform.

Here, the transverse crack is a crack that passes through (not necessarily) through the fiber reinforced resin layer 52a in the thickness direction, and the high-pressure tank eventually breaks due to the increase in the transverse crack. Microcracks are shorter in length than transverse cracks. Transverse cracks are formed by connecting a plurality of microcracks. The microcracks are generated from a state without cracks, while the transverse cracks are formed by connecting a plurality of microcracks, that is, because the formation processes of the two are different. It is considered that the generated AE wave also has a difference. The length of the transverse crack is often 0.1 mm or more due to the connection of a plurality of microcracks. The transverse crack is generated by connecting a plurality of microcracks, but it is considered that the transverse crack is also generated from a void in the reinforcing layer 52 as a starting point.

The conversion unit 122 sequentially performs time-frequency analysis on the AE waveform input from the extraction unit 121. In the present embodiment, the conversion unit 122 wavelet transforms the AE waveform to generate an image (scalogram) as shown in FIG. Here, the horizontal axis of the image (scalogram) shown in FIG. 6 indicates time, the vertical axis indicates frequency, and the color (density) indicates intensity. The wavelet transform is a frequency analysis using a wavelet function as a basis function. In the wavelet transform, unlike the Fourier transform, the frequency characteristics can be calculated while leaving the temporal information. The wavelet transform is defined by the following equation (1).

In equation (1), a is a scale parameter, b is a shift parameter, and Ψ (t) is a mother wavelet. Since the wavelet transform itself is a known method, a detailed description thereof will be omitted.

The classifier 123 uses the k-means method as a machine learning algorithm, and the output (image) of the conversion unit 122 is the first image corresponding to the first waveform derived from the transverse crack of the high-pressure tank 50 and the micro. It is classified into a second image corresponding to the second waveform derived from cracks, a third image corresponding to the third waveform derived from delamination, and a fourth image corresponding to the fourth waveform derived from fiber breakage. It is machine-learned in advance. Further, the classifier 123 outputs the classified first to fourth images to the counting unit 124. The machine learning method of the classifier 123 will be described later.

The counting unit 124 counts the numbers of the first image to the fourth image, respectively, and outputs the number to the output unit 125. The counting unit 124 may also be able to count the number of AE waveforms extracted by the extraction unit 121, if necessary.

The output unit 125 outputs the numbers and ratios of the first image to the fourth image to the display unit 22.

The pressurizing execution unit 126 detects the internal pressure of the high pressure tank 50 from the pressure signal from the pressure sensor 4. When the internal pressure of the high-pressure tank 50 reaches the upper limit of the inspection pressure, the pressurizing execution unit 126 stops driving the pump 2 and sends a valve drive signal for closing the valve 7 and opening the valve 9 to the valve opening / closing unit 127. Output. Further, the pressurization execution unit 126 outputs an output execution signal for outputting the number and ratio of each of the first image to the fourth image to the display unit 22 to the output unit 125.

When a valve drive signal is input from the pressurizing execution unit 126, the valve opening / closing unit 127 closes the valve 7 and opens the valve 9.

Next, a method of machine learning the classifier 123 will be described.

In this embodiment, as shown in FIG. 6, the classifier 123 is machined using the teacher data in which each image (scalogram) is associated with the damage mode (transverse crack, microcrack, delamination, fiber breakage). Let them learn. Here, since the moment when a crack or the like occurs cannot be visually captured, the first image corresponds to a transverse crack, the second image corresponds to a microcrack, and the third image corresponds to delamination. It is necessary to confirm that the fourth image corresponds to the fiber breakage.

First, a method of acquiring teacher data when performing machine learning will be described. When the pressure inside the high pressure tank 50 is increased, transverse cracks, macrocracks, delamination and fiber breakage occur in the reinforcing layer 52.

Therefore, the internal pressure of the high-pressure tank 50 for learning is increased to a predetermined pressure, and the images (first image to fourth image) obtained by wavelet transforming the plurality of AE waveforms obtained at that time and the pressure is reduced to a predetermined pressure. , The relationship with the number of each damage mode (transverse crack, microcrack, delamination and fiber breakage) measured by observing the cross section of the reinforcing layer 52 is confirmed. This is done for some internal pressures from a relatively low internal pressure (eg, several to several tens of MPa) to the destruction of the high pressure tank 50. By observing the cross section of the reinforcing layer 52, the high-pressure tank 50 for learning cannot be used. Therefore, it is necessary to prepare a plurality of high-pressure tanks 50 for learning.

Then, for each internal pressure, the number and ratio of each of the first to fourth images, and the number and ratio of each of the transverse cracks, microcracks, delamination, and fiber breakage generated in the reinforcing layer 52. By comparing, for example, the transverse crack corresponds to the first image, the microcrack corresponds to the second image, the delamination corresponds to the third image, and the fiber breakage corresponds to the fourth image. It can be confirmed that it is supported.

Further, as the internal pressure of the high pressure tank 50 increases, the rate of increase in the cumulative number of each of the first image to the fourth image (the rate of increase in the cumulative number with respect to the amount of pressure increase) increases, but in particular, the first image and the fourth image It can be confirmed that the rate of increase in the cumulative number of each image increases.

When confirming the correspondence between the first to fourth images and transverse cracks, microcracks, delamination, and fiber breakage, for example, a high-pressure tank 50 in which transverse cracks are likely to occur is manufactured, and the high-pressure tank is manufactured. By detecting the AE wave generated by pressurizing the 50 and observing the cross section of the high pressure tank 50, it is possible to easily confirm that the transverse crack corresponds to the first image. Since the transverse crack is mainly generated from the void formed inside the reinforcing layer 52, a high-pressure tank in which the transverse crack is likely to occur by forming many voids inside the reinforcing layer 52. You can get 50. Conversely, as will be described later, when a relatively large number of transverse cracks were detected when the high-pressure tank 50 was inspected using the tank inspection device 20, many voids were formed in the high-pressure tank 50. Turns out.

Similarly, it is also possible to manufacture a high-pressure tank 50 in which delamination and fiber breakage are likely to occur. For example, when the semi-cured fiber-reinforced resin layer 52a is wound around the liner 51 and cured, a high-pressure tank 50 in which delamination is likely to occur can be obtained, and the fiber-reinforced resin layer 52a and fibers made of a material other than carbon fiber can be obtained. When the reinforcing layer 52 is formed by using the fiber-reinforced resin layer 52a having a small diameter, a high-pressure tank 50 in which fiber breakage is likely to occur can be obtained. Since the number of microcracks generated is much larger than that of other damage modes, especially at a relatively low pressure (for example, several tens of MPa), it is not necessary to manufacture the high-pressure tank 50 in which microcracks are likely to occur. From the number of occurrences, it can be easily confirmed that the transverse crack and the first image correspond to each other.

As described above, it can be confirmed that the first image corresponds to the transverse crack, the second image corresponds to the microcrack, the third image corresponds to the delamination, and the fourth image corresponds to the fiber breakage. It can be said that. Then, as shown in FIG. 6, data corresponding to each image corresponding to the first image (transverse crack), the second image (microcrack), the third image (delamination), and the fourth image (fiber breakage) are obtained. It can be teacher data.

Next, the classifier 123 was machine-learned using the teacher data. Specifically, 70 to 80% of the hundreds of teacher data are used as training data, and the image is used for delamination between the first image derived from the transverse crack and the second image derived from the microcrack. The classifier 123 is machine-learned to classify into a corresponding third image and a fourth image corresponding to fiber breakage. Then, when the images are classified into the first image to the fourth image by using the rest of the hundreds of teacher data as the evaluation data, the correct answer rate of the classifier 123 is 70% or more, and the classification accuracy is sufficient. Turns out to be high. As described above, the classifier 123 is machine-learned.

Next, a pressure inspection method for the high pressure tank 50 using the tank inspection device 20 of the present embodiment will be described with reference to FIG. 7. In this pressure inspection method, when designing the high-pressure tank 50, various manufacturing conditions (fibers) when the material (fiber, resin material) constituting the reinforcing layer 52 and the fiber reinforced resin layer 52a are wound around the liner 51. Evaluate the performance of the high-pressure tank 50 manufactured by changing the winding angle of the reinforced resin layer 52a, the number of times the fiber reinforced resin layer 52a (fiber bundle) is wound at the same time (hereinafter, also referred to as the number of threads to be fed), the curing conditions, and the like. It is used when.

In step S1, when the operator operates the operation unit 23 of the tank inspection device 20, the pressure inside the high-pressure tank 50 is increased by the tank inspection device 20. At this time, the valve 7 is open and the valve 9 is closed. The internal pressure of the high pressure tank 50 is increased at a substantially constant speed.

In step S2, when transverse cracks, macrocracks, delamination and fiber breakage occur in the reinforcing layer 52 as the internal pressure of the high pressure tank 50 increases, the AE sensor 3 outputs a waveform including the AE waveform to the extraction unit 121. do. The extraction unit 121 sequentially extracts the AE waveform from the output waveform of the AE sensor 3 and outputs it to the conversion unit 122.

In step S3, the conversion unit 122 wavelet transforms the AE waveform input from the extraction unit 121 to sequentially generate an image (scalogram). Further, the conversion unit 122 outputs the generated image to the classifier 123.

In step S4, the classifier 123 sequentially outputs (images) the output (image) of the conversion unit 122 to the first image corresponding to the first waveform, the second image corresponding to the second waveform, and the third image corresponding to the third waveform. It is classified into an image and a fourth image corresponding to the fourth waveform. Further, the classifier 123 outputs the classified first to fourth images to the counting unit 124.

In step S5, the counting unit 124 counts the number of first to fourth images sequentially input from the classifier 123 (adding 1 to N). Further, the counting unit 124 outputs the counted numbers of the first image to the fourth image to the output unit 125.

In step S6, the pressurizing execution unit 126 determines whether or not the internal pressure of the high pressure tank 50 is equal to or higher than the upper limit value of the inspection pressure based on the output of the pressure sensor 4. If the internal pressure of the high pressure tank 50 is less than the upper limit of the inspection pressure, the process returns to step S2.

On the other hand, in step S6, when the internal pressure of the high pressure tank 50 is equal to or higher than the upper limit of the inspection pressure, the pump 2 is stopped to stop the pressurization of the high pressure tank 50, and the valve drive signal and the output execution signal are opened and closed. Output to unit 127 and output unit 125, respectively. As a result, the valve opening / closing unit 127 closes the valve 7 and opens the valve 9. Therefore, the supply of water to the high-pressure tank 50 is stopped, and the water in the high-pressure tank 50 is discharged to the outside through the discharge pipe 8. As a result, the internal pressure of the high pressure tank 50 decreases. Further, the output unit 125 outputs the number and ratio of each of the first image to the fourth image to the display unit 22.

In step S7, the display unit 22 displays the respective numbers and proportions of the first to fourth images, that is, the respective numbers and proportions of transverse cracks, microcracks, delaminations, and fiber breaks.

Then, the pressurization inspection of the high pressure tank 50 is completed.

The designer of the high-pressure tank 50 can evaluate the performance of the high-pressure tank 50 from the result displayed on the display unit 22 without cutting the high-pressure tank 50 and observing the cross section. For example, it is known that the generation of voids in the reinforcing layer 52 is largely due to the winding angle of the fiber reinforced resin layer 52a and the number of threads fed. Therefore, by performing the above-mentioned pressure inspection using a plurality of high-pressure tanks 50 having different winding angles and the number of yarn feeds of the fiber-reinforced resin layer 52a, manufacturing conditions with less occurrence of transverse cracks (optimum of the fiber-reinforced resin layer 52a). The winding angle and the optimum number of yarns to be fed) can be specified. For example, if the number of yarns to be fed is increased from one to nine, the number of voids generated at both ends (dome-shaped portion) of the high-pressure tank 50 tends to increase. Therefore, it is effective to adopt the inspection method of the present embodiment when designing to increase the number of yarns to be fed in order to shorten the manufacturing time. Further, the high-pressure tank 50 having many voids has the same initial breaking strength (internal pressure at the time of breaking) as the high-pressure tank 50 having few voids, but the high pressure when the gas is repeatedly filled and released into the inside. The number of repetitions until the tank 50 is destroyed is smaller than that of the high-pressure tank 50 having few voids. That is, the high-pressure tank 50 having many voids has a low cycle strength. Therefore, it is particularly effective to design the high-pressure tank 50 with less generation of voids by using the inspection method of the present embodiment.

All the AE waveforms extracted by the extraction unit 121 may be converted into images by the conversion unit 122 and classified into the first image to the fourth image by the classifier 123. However, since tens of thousands of AE waveforms are extracted every time the internal pressure of the high-pressure tank 50 increases by 5 MPa, it is not easy to convert and classify all AE waveforms. Therefore, for example, hundreds of AE waveforms extracted by the extraction unit 121 are selected, converted and classified for hundreds of AE waveforms, the ratio of the first image to the fourth image is calculated, and the ratio of each image is calculated. And the total number of AE waveforms extracted by the extraction unit 121, the cumulative number of each image may be calculated.

Further, when three or more AE sensors 3 are used, the position where the AE wave is generated in the high pressure tank 50 is detected by using a known technique based on the time difference when the AE wave reaches the AE sensor 3. It is possible. By detecting the generation position of the AE wave using the tank inspection device 20 and displaying the generation position on the display unit 22, for example, the position where many voids are generated can be specified. Can be more optimized.

In the present embodiment, as described above, a plurality of AE waveforms are extracted from the output waveform of the AE sensor 3, and the plurality of AE waveforms are divided into a first waveform derived from a transverse crack and a second waveform derived from a microcrack. , And a third waveform derived from delamination and a fourth waveform derived from fiber breakage, and the numbers of the first to fourth waveforms classified are counted. As a result, the damage mode (transverse crack, microcrack, delamination, fiber breakage) of the high pressure tank 50 and the number of occurrences thereof can be confirmed based on the counted result, so that the state of the reinforcing layer 52 can be confirmed. Therefore, it is not necessary to cut the high pressure tank 50 and observe the cross section. Therefore, the time required for checking the state of the reinforcing layer 52 of the high-pressure tank 50 can be shortened. Further, unlike the case of observing the cross section, the state of occurrence of cracks, delamination and the like does not differ depending on the cutting position, so that the state of the reinforcing layer 52 can be confirmed accurately. As a result, the time required for designing the high-pressure tank 50 can be significantly reduced.

Further, as described above, by wavelet transforming the extracted AE waveform, an image (scalogram) showing the time change of the frequency component is generated. Then, by classifying the generated plurality of images with the classifier 123, the first image to the fourth image can be easily classified, so that the damage mode of the high-pressure tank 50 and the number of occurrences thereof can be easily confirmed. be able to.

(Second Embodiment)
In the second embodiment, unlike the first embodiment, a case where the inside of the high pressure tank 50 is pressurized by filling the inside of the high pressure tank 50 with gas will be described.

The pump 2 fills the inside of the high-pressure tank 50 with gas (here, nitrogen) to pressurize the inside of the high-pressure tank 50. Nitrogen is filled in the high pressure tank 50 via the supply pipe 6 and discharged to the outside of the high pressure tank 50 through the discharge pipe 8. Nitrogen may be discharged into the examination chamber via the discharge pipe 8. The gas filled in the high-pressure tank 50 is not limited to nitrogen, and for example, helium, air, or a mixed gas of helium and nitrogen may be used.

In the present embodiment, since the inside of the high-pressure tank 50 is filled with gas (nitrogen in this case), the AE wave (vibration) generated when AE is generated in the high-pressure tank 50 propagates through the reinforcing layer 52 and AE. It is detected by the sensor 3. Therefore, the output waveform of the AE sensor 3 becomes a waveform in which a large number of waveforms as shown in FIG. 8 are connected. The region B in FIG. 8 corresponds to the region A1 (vibration propagating in the reinforcing layer 52) in FIG. 5, and the AE waveform is amplified.

In the present embodiment, as shown in FIG. 9, the control unit 21 of the tank inspection device 20 includes an extraction unit 121, a classifier 123, a count unit 124, an output unit 125, a pressurization execution unit 126, and a valve opening / closing unit 127. On the other hand, the conversion unit 122 is not included.

The output waveform of the AE sensor 3 is input to the extraction unit 121, and when the AE occurs in the high-pressure tank 50 and the amplitude becomes equal to or higher than a predetermined threshold value (L), the extraction unit 121 vibrates a mass (region B) by 1. Extract as two AE waveforms. Since this AE waveform does not include vibration propagating in water, that is, a wave (vibration) propagating at the interface between water and the liner 51, a wave (vibration) propagating in the reinforcing layer 52, and a wave (vibration) propagating in water. Unlike the first embodiment, the AE waveform (output waveform of the AE sensor 3) is accurately classified into the first waveform to the fourth waveform even if the conversion unit 122 is not provided. It is possible to do.

AE waveforms are sequentially input to the classifier 123 from the extraction unit 121. The classifier 123 converts the output (AE waveform) of the extraction unit 121 into a first waveform derived from a transverse crack, a second waveform derived from a microcrack, a third waveform derived from delamination, and fiber breakage. It has been machine-learned in advance so as to classify it as a fourth waveform from which it is derived. Further, the classifier 123 outputs the classified first waveform to the fourth waveform to the counting unit 124.

The counting unit 124 counts the numbers of the first waveform to the fourth waveform, and outputs the result to the output unit 125.

The output unit 125 outputs the number and ratio of the first waveform to the fourth waveform to the display unit 22.

Other configurations of the second embodiment are the same as those of the first embodiment. The features used for the teacher data when the classifier 123 is machine-learned include the maximum amplitude (maximum intensity) of each AE waveform, the rise time, the number of times the signal exceeds the threshold, the waveform duration, and the energy. .. For example, the AE waveform may be classified into four based on two or more features (maximum amplitude, waveform duration, etc.). It is also possible to configure the tank inspection device 20 to select the feature amount used for classification by itself.

Next, a pressure inspection method for the high pressure tank 50 using the tank inspection device 20 will be described with reference to FIG.

In the present embodiment, steps S1 and S2 are the same as those in the first embodiment. However, in step S2, the extraction unit 121 outputs the AE waveform extracted from the output waveform of the AE sensor 3 to the classifier 123. In this embodiment, step S3 is not provided.

In step S4, the classifier 123 sequentially classifies the output (AE waveform) of the extraction unit 121 into the first waveform to the fourth waveform. Further, the classifier 123 outputs the classified first waveform to the fourth waveform to the counting unit 124.

In step S5, the counting unit 124 counts the respective numbers of the first waveform to the fourth waveform, and outputs the result to the output unit 125.

Step S6 is the same as that of the first embodiment. In step S7, the display unit 22 displays the number and ratio of the first waveform to the fourth waveform, that is, the number and ratio of transverse cracks, microcracks, delamination, and fiber breakage, respectively. ..

The other pressurization inspection method and other effects of the present embodiment are the same as those of the first embodiment.

It should be noted that the embodiments disclosed this time are exemplary in all respects and are not considered to be restrictive. The scope of the present invention is shown by the scope of claims rather than the description of the above-described embodiment, and further includes all modifications within the meaning and scope equivalent to the scope of claims.

For example, in the above embodiment, an example of classifying a plurality of AE waveforms extracted from the output waveform of the AE sensor into waveforms caused by any of transverse cracks, microcracks, delamination, and fiber breakage has been shown. The invention is not limited to this. For example, when the reinforcing layer has an under-cured (burnt) portion, the AE wave is attenuated when the AE wave propagates through the under-cured portion, or the frequency of the AE wave becomes low. Therefore, if the amplitude of the AE waveform becomes small (or the number of AE waveforms extracted becomes small) or the frequency of the AE waveform becomes lower than the frequency that is normally detected, the reinforcing layer is insufficiently cured. It is possible to detect the presence of. Further, as described above, when three or more AE sensors 3 are used, it is possible to detect the position of the insufficiently cured portion of the reinforcing layer. Therefore, it may be configured to detect insufficient curing of the reinforcing layer based on the AE waveform. With this configuration, it is possible to set the curing conditions of the reinforcing layer (set temperature, heater position, hot air application method, etc.) to the optimum conditions, and to detect a heater failure.

Further, in the above embodiment, there is an example in which a plurality of AE waveforms are extracted from the output waveform of the AE sensor, the AE waveforms are classified into the first waveform to the fourth waveform, and the respective numbers are output while pressurizing the high pressure tank. As shown, the present invention is not limited to this, and pressurization, classification, and output of the high-pressure tank may be performed separately. Specifically, the AE wave is detected by the AE sensor while the high pressure tank is pressurized to the upper limit of the inspection pressure, the output waveform of the AE sensor is stored, and the high pressure tank is depressurized. After that, a plurality of AE waveforms may be extracted from the stored output waveforms of the AE sensor, the AE waveforms may be classified into the first waveform to the fourth waveform, and the respective numbers may be output.

Further, in the first embodiment described above, an example of wavelet transforming an AE waveform has been described, but the present invention is not limited to this, and the AE waveform can be transformed by using a method such as a fast Fourier transform, a short-time Fourier transform, and a Wigner distribution. It may be converted.

Further, in the second embodiment described above, an example in which the AE waveform is not converted is shown, but the present invention is not limited to this. Even when the high-pressure tank is pressurized by filling with gas, the AE waveform may be classified after being converted by wavelet transform or the like. With this configuration, it is possible to classify the first image (or the first waveform) to the fourth image (or the fourth waveform) with higher accuracy.

Further, in the above embodiment, an example in which the k-means method is used as the machine learning algorithm of the classifier has been described, but the present invention is not limited to this. For example, other machine learning algorithms such as support vector machine VAE (Variational Auto-Encoder), CNN (Convolutional Neural Network), GAN (Generative Adversarial Network), Bayesian filter, or isolation forest may be used for classification.

3: AE sensor, 20: tank inspection device, 50: high pressure tank, 51: liner, 52a: fiber reinforced resin layer, 52: reinforcing layer, 121: extraction unit, 123: classifier, 124: count unit, 125: output Department

Claims (1)

A tank inspection device that inspects a high-pressure tank including a liner and a reinforcing layer made of a fiber-reinforced resin that covers the outer surface of the liner. A plurality of AE waveforms are extracted from the output waveform of the AE sensor that detects the AE wave generated in the high pressure tank, and the high pressure tank is inspected based on the extracted plurality of AE waveforms.
The reinforcing layer is composed of a plurality of fiber reinforced resin layers.
The tank inspection device is
An extraction unit that extracts the plurality of AE waveforms from the output waveforms of the AE sensor that detects the AE waves generated in the high-pressure tank, and
The plurality of AE waveforms extracted by the extraction unit are derived from a first waveform derived from a transverse crack that passes through the fiber reinforced resin layer in the thickness direction and a microcrack that does not pass through the fiber reinforced resin layer in the thickness direction. A classifier machine-learned to classify into a second waveform, a third waveform derived from delamination of the fiber reinforced resin layer, and a fourth waveform derived from fiber breakage.
A counting unit that counts the respective numbers of the first waveform to the fourth waveform classified by the classifier, and
An output unit that outputs the result counted by the count unit, and an output unit.
A tank inspection device comprising a computing device having the above.

Images