Fractographic analysis device and fractographic analysis method

Abstract

The present invention provides a fracture analysis device and method that enable unskilled personnel to perform the same fracture analysis as skilled personnel. The device and method are characterized by comprising a microscopic region observation unit, comprising: a first unit for calculating the accuracy of a failure mode for a first observation image of an object to be analyzed, obtained by a microscopic region imaging unit; a second unit for calculating observation conditions that can improve the accuracy of the failure mode; and a third unit for adjusting the imaging conditions of the microscopic region imaging unit and outputting the failure mode accuracy and the observation conditions that can improve the accuracy of the failure mode to a visualization unit.

Description

Technical Field

[0001] The invention relates to a fracture analysis device and a fracture analysis method for performing component fracture analysis. Background Art

[0002] Fracture analysis is a common method for identifying the cause of component failure. In fracture analysis, macroscopic observations such as visual observation of the fracture or microscopic observations using an electron microscope are performed to determine the cause and process of the fracture. Fracture analysis can capture the characteristic shape and texture corresponding to the failure mode that appears on the fracture surface. In macroscopic observation, the shape of the fracture and the macroscopic characteristic texture are captured to estimate the failure mode of each part of the fracture and determine the starting point of the fracture. In microscopic observation, the failure mode is determined and the failure process is estimated in detail based on the search for the characteristic texture corresponding to the failure mode in each part of the fracture and the change in the characteristic texture of the fracture corresponding to the distance to the starting point of the fracture. In addition, the fracture image obtained for fracture analysis can be used to identify the failure process and can also be used as objective evidence to show others the cause of the failure.

[0003] Patent Document 1 discloses a method and an observation device for classifying the types of fractures by analyzing the color depth of observed fractures.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Laid-Open No. 2018-77719 Summary of the Invention

[0007] Technical problem to be solved by the invention

[0008] Fracture analysis relies heavily on the judgment of skilled personnel based on implicit knowledge, but in recent years, as skilled personnel are aging, the demand for assistance / automation of fracture analysis is gradually increasing. In order to reproduce the series of fracture observation processes performed by skilled personnel, it is required not only to be able to classify the failure mode based on the microscopic observation image, but also to be able to determine the observation position in the microscopic observation based on the failure mode of each part of the fracture and the estimated results of the fracture starting point obtained by macroscopic observation. In addition, in microscopic observation, it can be considered that the operation performed by skilled personnel is to search for the part that most clearly shows the characteristic texture corresponding to the failure mode, and obtain an image suitable for use as evidence to represent the failure mode based on the adjustment of shooting conditions such as the observation magnification. Therefore, it can be considered that a technology is needed to assist in the optimization of the shooting position and shooting conditions.

[0009] However, until now, there has been little research on specific methods for optimizing the imaging conditions for microscopic observation images.

[0010] The present invention is made to solve such problems and provides a fractographic analysis device and a fractographic analysis method that assist fractographic analysis so that unskilled personnel can perform the same fractographic analysis as skilled personnel.

[0011] Technical means to solve the problem

[0012] The fracture analysis device of the present invention for achieving the above-mentioned purpose is characterized in that it includes a microscopic area observation part, which includes: a first unit, which calculates the accuracy of the destruction mode for a first observation image of the object to be analyzed obtained by the camera unit of the microscopic area; a second unit, which calculates the observation conditions that can improve the accuracy of the destruction mode; and a third unit, which adjusts the shooting conditions of the camera unit of the microscopic area, and outputs the accuracy of the destruction mode and the observation conditions that can improve the accuracy of the destruction mode to the visualization unit.

[0013] The fracture analysis method of the present invention is characterized in that, for the first observation image of the object to be analyzed obtained by the camera unit of the microscopic area, the accuracy of the destruction mode is calculated, the observation conditions that can improve the accuracy of the destruction mode are calculated, the shooting conditions of the camera unit of the microscopic area are adjusted, and the accuracy of the destruction mode and the observation conditions that can improve the accuracy of the destruction mode are displayed, wherein the accuracy of the destruction mode is obtained for multiple first observation images with different shooting positions, and the shooting position on the side with higher accuracy is represented as the observation condition that can improve the accuracy of the destruction mode.

[0014] The fracture analysis method of the present invention is characterized in that, for a first observation image of the object to be analyzed obtained by a camera unit of a microscopic area, the accuracy of the destruction mode is calculated, the observation conditions that can improve the accuracy of the destruction mode are calculated, the shooting conditions of the camera unit of the microscopic area are adjusted, and the accuracy of the destruction mode and the observation conditions that can improve the accuracy of the destruction mode are displayed, the first observation image is divided into multiple areas, the accuracy of the destruction mode is obtained for each divided area, and the shooting position of the divided area with the highest accuracy is represented as the observation condition that can improve the accuracy of the destruction mode.

[0015] Effects of the Invention

[0016] The present invention allows for more precise microscopic capture of characteristic textures corresponding to failure modes, enabling the acquisition of high-quality images that serve as objective evidence for failure mechanisms. Consequently, the use of the fracture analysis device of the present invention enables even unskilled personnel to perform high-speed, accurate, and convincing fracture analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a diagram showing a configuration example of an embodiment of a fracture analysis device according to the present invention.

[0018] Figure 2 This is a diagram showing an example of a screen configuration in the visualization unit 1 .

[0019] Figure 3 1 and 2 are diagrams showing an example of processing performed by the calculation unit 6 for the accuracy of the observed image fracture pattern.

[0020] Figure 4 Schematically shows a series of processes in the microscopic region observation section 101 .

[0021] Figure 5 Schematically shows another series of processing in the microscopic region observation section 101 . DETAILED DESCRIPTION

[0022] The following is an explanation of preferred embodiments of the present invention using the accompanying drawings. The following are merely examples and are not intended to limit the invention to the specific embodiments described below. The invention itself can be modified in various ways within the scope of its technical solution.

[0023] Example

[0024] The following is based on Figures 1 to 5 Specific embodiments of the present invention are described in detail.

[0025] Figure 1 This figure shows an example structure of an embodiment of a fracture analysis device according to the present invention. The fracture analysis device 100 according to the embodiment of the present invention is configured using a computer and is broadly divided into a macro observation unit 102 for observing macroscopic areas and a microscopic area observation unit 101 for observing microscopic areas, based on their processing functions.

[0026] The fracture analysis device 100 is equipped with a microscopic imaging unit 2 and a macroscopic imaging unit 12 as input units, and also includes a keyboard or other suitable input unit (not shown). The fracture analysis device 100 can also be connected to the outside via communication. The imaging units 2 and 12 capture the overall image and surface image of the analyte 3 placed on the mounting table 4.

[0027] Furthermore, the fracture analysis device 100 is provided with a visualization unit 1 such as a monitor as its output unit. As the display content of the visualization unit 1, the observation image of the macroscopic area, the observation image of the microscopic area, the observation conditions when observing these areas, the processing judgment results, etc. can be displayed. Figure 2 The fracture analysis device 100 may also be configured to include the input unit and the output unit.

[0028] In the following description of the present invention, the final output, that is, the display content in the visualization unit 1 is first described, and then the microscopic area observation is sequentially described.

[0029] Figure 2 The following shows an example of the screen configuration of the visualization unit 1. The display screen 20 of the visualization unit 1 is composed of, for example, four small display regions R. The small region R1 is a whole image of the entire analyte 3 obtained by the macro-region imaging unit 12 and is a display screen region for the macroscopic observation image.

[0030] In addition, the area of ​​the microscopic observation position 502 at this time is displayed in a frame in the display content of the small area R1 showing the entirety. This makes it possible to clearly identify the position where the microscopic observation image is obtained on the macroscopic fracture. The microscopic observation image at this time is displayed in the small area R2.

[0031] The small area R3 displays the recommended observation conditions for microscopic observation as the processing result of the fracture analysis device 100, and teaches the observer the contents of the operation to be performed, for example, teaching the observer to increase the observation magnification.

[0032] Small area R4 displays the calculated fracture pattern accuracy of the observed image, obtained as a result of processing by the fracture analysis device 100. In the illustrated example, the fracture diagnosis results show that the fatigue fracture accuracy (confidence level) at the current location is 0.98, and the ductile fracture accuracy is 0.01. The accuracy values ​​can be displayed as either the value at that location or the difference from the accuracy value at a recommended observation location that improves failure mode accuracy.

[0033] Next, explain Figure 1 The processing of the macro-region observation unit 102 is premised on the fact that the macro-region imaging unit 12, which is the input unit of the macro-region observation unit 102, is primarily used to capture images of the entire fracture surface of the analyte 3. Unlike the micro-region imaging unit 2, the macro-region imaging unit 12 is not particularly preferably a unit that uses an electron beam and may also be an optical camera or the like.

[0034] The stage 4 is used to place the analyte 3. The analyte 3 has a fractured surface and is secured to the stage 4 using tape or an adhesive. The stage 4 can be moved vertically, vertically, and horizontally. This movement allows for adjustment of the observation position and for focusing the image of the microscopic region obtained by the imaging unit 2.

[0035] The macro-region observation unit 102 processes the entire image captured by the macro-region imaging unit 12. Figure 1In the small region R1 of FIG. 1 , an overall image of the entire analyte 3 is formed by the imaging unit 12 in the macroscopic region. When the overall image is displayed, the region of the microscopic observation position 502 is also displayed.

[0036] Next, explain Figure 1 The processing of the mesoscopic region observation unit 101 presupposes that the microscopic region imaging unit 2, which serves as the input unit of the microscopic region observation unit 101, primarily observes the microstructure of the fracture surface at a magnification of 100x or greater. Imaging units using electron beams, such as electron microscopes, are particularly preferred. This is because electron beams have a long depth of focus, making them suitable for observing analytes with strongly concave and convex surfaces, such as fracture surfaces, and with inclined observation surfaces.

[0037] The microscopic region observation section 101 includes a microscopic region imaging unit 2 , a microscopic observation image storage unit 5 , an observation image fracture pattern accuracy calculation unit 6 , an accuracy-enhancing observation condition calculation unit 7 , and an imaging position / condition adjustment unit 8 .

[0038] The microscopic region observation unit 101 uses the microscopic region imaging unit 2 to capture the surface of the analyte 3 placed on the mounting table 4, and stores the captured image in the microscopic observation image storage unit 5. The observed image stored in the microscopic observation image storage unit 5 is then processed by the observed image fracture pattern accuracy calculation unit 6 and the accuracy-enhancing observation condition calculation unit 7, and displayed as an image of microscopic region information on a monitor or other device that visualizes recommended observation conditions. Furthermore, the imaging position / condition adjustment unit 8 adjusts the relative position and angle of the microscopic region in the imaging unit 2 as appropriate to observe a different surface area of ​​the analyte 3, and observation continues.

[0039] For the composition Figure 1 These devices and processing functions of the microscopic region observation section 101 will be described in more detail.

[0040] First, the microscopic observation image storage unit 5 is used to store the observation image of the fracture surface obtained by the microscopic region imaging unit 2, and uses a recording medium such as a hard disk or solid state drive. Alternatively, it can be stored in the cloud.

[0041] The calculation unit 6 for the accuracy of the observed image fracture pattern uses a convolutional neural network or the like that has learned the relationship between the fracture image and the failure pattern.

[0042] The fracture pattern accuracy calculation unit 6 outputs the accuracy of each failure mode as a classification target by inputting the fracture image. Figure 2The input image (displayed in small area R3) is input to the observation image failure mode accuracy calculation unit 6, which outputs the accuracy for various failure modes such as ductile failure and fatigue failure. The input image and the accuracy of each failure mode are recorded separately. To reduce the computational load, the image size can be reduced when inputting the microscopic observation image to the observation image fracture mode accuracy calculation unit 6.

[0043] Figure 3 1 and 2 are diagrams showing an example of processing performed by the calculation unit 6 for the accuracy of the observed image fracture pattern. Figure 3 The object of processing is the observation image 201 of the fracture surface where the crack is generated, developed and finally reaches ductile failure due to fatigue. Figure 3 The observation image 201 shown is Figure 2 The image of the whole image is displayed in the small area R1 of the observation image 201. Figure 2 As shown in the example of the small area R1, the overall image including the fatigue fracture area 202, the ductile fracture area 203, and the fracture starting point 204 is displayed. Figure 2 In the microscopic region observation section 101, the image 207 of the local region 502 is observed as an object.

[0044] Here, the fatigue fracture region 202 is observed by the microscopic region imaging unit 2. In the microscopic region observation image 207, a wave-like texture called striation 208, which is more microscopic than the beach mark 205, is observed. The microscopic region fracture image 207 is input to the observation image fracture pattern accuracy calculation unit 6 composed of a learned neural network. Figure 2 In the display area R4 of , the accuracy of the fracture image 207 in the microscopic region with respect to various failure mechanisms such as fatigue failure and ductile failure is output as an output.

[0045] The accuracy-enhancing observation condition calculation unit 7 applies the observed image fracture pattern accuracy calculation unit 6 to the microscopic region observation image obtained by slightly changing the observation position and conditions. Based on the change in the value of the mode with the highest accuracy, the unit searches for the observation condition that maximizes the accuracy of each mode. Alternatively, the microscopic region imaging unit 2 can be used to observe at a low magnification of approximately 1 / 10 of the normal magnification, dividing the observed image into, for example, 16 sections. The accuracy of each section is then compared, and the section with the highest accuracy is then observed at a higher magnification. It is believed that the higher the accuracy, the more clearly the characteristic texture of the fracture corresponding to the failure mode can be visualized, enabling highly accurate and convincing fracture diagnosis.

[0046] The calculation unit 7 of the observation condition is used to improve the accuracy and displays the processing result on the Figure 2 In the display area R3. Figure 2 In the display area R3, the operation direction and magnification are displayed as operation conditions for improving the accuracy of observation.

[0047] The imaging position / condition adjustment unit 8 is used to adjust the position of the mounting stage 4, the acceleration voltage, observation magnification, brightness, and contrast of the imaging unit 2 in the microscopic area, and the observation magnification of the imaging unit 102 in the macroscopic area. The imaging position and conditions can be set based on the values ​​of the various adjustment parameters specified by the observer via the control software, but it is also possible to adopt a method in which the imaging position and conditions are automatically adjusted based on the results obtained by the accuracy-enhancing observation condition calculation unit 7.

[0048] In addition to displaying the observation field image of the microscopic region captured by the imaging unit 2, the visualization unit 1 also displays the recommended observation position movement direction for improving diagnostic accuracy, as calculated by the accuracy-enhancing observation condition calculation unit 7, and guidelines for changing the parameters defining the imaging conditions. This enables unskilled personnel to perform high-speed, high-precision fracture diagnosis. While a liquid crystal display is typically used for the visualization unit 1, any display capable of visualizing the recommended observation conditions will suffice.

[0049] Next, for Figure 1 The series of processing contents in the fracture analysis device shown are used Figure 4 、 Figure 5 Provide explanation.

[0050] Figure 4 Schematically shows a series of processes in the microscopic region observation section 101. Figure 4 In the processing of the present invention, in processing step S1, the display content of small area R1 representing the entire fracture morphology of the analyte 3 is used as the observation object. In processing steps S2 and S3, the electron microscope microscopic (high-magnification) fracture images displayed in the display area of ​​small area R2 are obtained under observation conditions A and B. These images are respectively subjected to the neural network processing for fracture diagnosis in processing steps S4 and S5.

[0051] According to the diagnosis, when the observation condition is B, an area is included where the characteristic texture cannot be observed due to dirt, damage, etc. As a result, in processing steps S6 and S7, the accuracy of the fatigue fracture is calculated to be 0.95 under observation condition A and 0.70 under observation condition B, respectively, and they are displayed in the small area R4.

[0052] Next, in processing step S8 , it is determined that the recommendation is to change the observation condition from observation condition B to observation condition A, and in processing step S9 , this is displayed in the small region R3 .

[0053] The reason for Figure 4 The purpose of this series of processing is to assist observation and to capture fracture images that clearly display the characteristic textures inherent to each fracture pattern. In this case, the more clearly the characteristic texture is displayed in the image, the more powerful evidence it can provide for the failure mechanism.

[0054] In addition, in the implementation Figure 4 When the fracture diagnosis neural network (with the observation image as input and the accuracy of each failure mode as output) that has learned the characteristic texture corresponding to the fracture pattern is prepared in advance, it is used as described below.

[0055] First, images observed under a microscopic (high-magnification) electron microscope before and after changes in observation conditions (magnification, observation position, etc.) are input into the fracture diagnosis neural network. Images with more pronounced characteristic textures have higher accuracy output from the neural network, indicating that the observation conditions are appropriate. Therefore, based on the accuracy comparison results, changing the observation conditions in the direction of increasing accuracy can produce fracture images with more pronounced characteristic textures.

[0056] On this basis, by informing the observer of the observation conditions in the direction in which the accuracy can be improved as recommended observation conditions, it is possible to assist in obtaining an image that clearly shows the characteristic texture.

[0057] Figure 5 Schematically shows another series of processing in the microscopic region observation section 101. Figure 5 In the processing step S11, the display content of small area R1, representing the entire fracture morphology of the analyte 3, is used as the observation object. In step S12, a microscopic fracture image is obtained as an electron microscope image displayed in the display area of ​​small area R2, using a lower magnification than that used in conventional fracture observation using an electron microscope. This allows the microscopic fracture image to be obtained over a larger area, allowing for the observation of more areas where characteristic textures are not visible due to dirt, damage, etc. Figure 5 In , an image of a region where no characteristic texture is observed is initially obtained.

[0058] On this basis, the region of the image obtained at low magnification is segmented into multiple parts in processing step S13. Here, it is assumed that the observed image is segmented into regions A, B, C, and D. For each of these segmented images, the neural network processing for fracture diagnosis in processing step S14 is performed.

[0059] The diagnosis of multiple split images in step S15 yields the following evaluation results: When observing image A, which is free of dirt and damage, the fatigue fracture accuracy is 0.99; when observing image B, which is free of dirt and damage, the fatigue fracture accuracy is 0.6; when observing image C, which is free of dirt and damage, the fatigue fracture accuracy is 0.7; and when observing image D, which is free of dirt and damage, the fatigue fracture accuracy is 0.8. Based on these accuracy evaluation results, in step S15, image A exhibits the highest accuracy for fatigue fracture, making it the most suitable recommended observation position. In step S16, a message is displayed in small area R4.

[0060] The reason for Figure 5 The purpose of this series of processing is to assist observation and to capture fracture images that clearly display the characteristic textures inherent to each fracture pattern. In this case, the more clearly the characteristic texture is displayed in the image, the more powerful evidence it can provide for the failure mechanism.

[0061] exist Figure 5 In a series of processing, an observation image of the fracture is obtained at a lower magnification than when the fracture is usually observed using an electron microscope, the observation image is divided into equal parts, and the divided images are input into the neural network to calculate the accuracy.

[0062] In the equally divided image, the area with the highest accuracy is most likely to represent the characteristic texture corresponding to the fracture pattern, and is therefore more likely to produce an image in which the characteristic texture is clearly displayed. Therefore, the position of the area with the highest accuracy in the equally divided image is provided to the observer as the recommended observation position, thereby assisting in obtaining an image in which the characteristic texture is clearly displayed.

[0063] Description of Reference Numerals

[0064] 1: Visualization unit, 2: Camera unit for microscopic area, 3: Analyte, 4: Mounting table, 5: Microscopic observation image storage unit, 6: Calculation unit for observation image fracture pattern accuracy, 7: Calculation unit for observation conditions for improving accuracy, 8: Shooting position / condition adjustment unit, 12: Camera unit for macroscopic area, 20: Display screen of visualization unit, 202: Fatigue fracture area, 203: Ductile fracture area, 204: Fracture starting point, 205: Beach pattern, 206: Acquisition position of microscopic observation image, 207: Microscopic observation image, 208: Striation, 209: Calculation results of the accuracy of each failure mode obtained by the calculation unit for observation image fracture pattern accuracy, 502: Acquisition position of microscopic observation image on macroscopic fracture.

Claims

1. A fracture analysis device, characterized in that: The microscopic region observation section includes: a first unit for calculating the accuracy of a destruction mode with respect to a first observation image of the analyte obtained by an imaging unit of a microscopic region; A second unit that calculates observation conditions including a magnification and an observation position that can improve the accuracy of the failure mode; and a third unit for adjusting a photographing condition of the photographing unit of the microscopic area, The accuracy of the failure mode and the observation conditions that can improve the accuracy of the failure mode are output to a visualization unit, and The first unit includes a neural network, which processes observation images including observation images of the fracture surface of the object being analyzed, including the observation images of the crack generation, development and ultimate destruction, and learns the relationship between the observation images and the destruction mode, wherein the first unit uses the learned neural network to obtain the accuracy of the first observation image ultimately reaching destruction for each of the destruction modes. The second unit obtains an observation condition that provides a higher accuracy among the accuracies of a plurality of destruction modes from the first unit when the observation condition is changed, with respect to the first observation image obtained by the imaging unit of the microscopic region of the analyte.

2. The fracture analysis device according to claim 1, wherein: The observation conditions that can improve the accuracy of the failure mode are expressed as adjustment contents in the third unit.

3. The fracture analysis device according to claim 1 or 2, wherein: The accuracy of the failure mode is expressed as the accuracy at the position where the observation image is acquired, or as the difference from the accuracy under observation conditions that can improve the accuracy of the failure mode.

4. A fracture analysis method, characterized in that: For a first observation image of the analyte obtained by the imaging unit of the microscopic region, the accuracy of the destruction mode is calculated, observation conditions including magnification and observation position that can improve the accuracy of the destruction mode are calculated, shooting conditions of the imaging unit of the microscopic region are adjusted, and the accuracy of the destruction mode and the observation conditions that can improve the accuracy of the destruction mode are displayed. Here, a learned neural network is used to process observation images including observation images of the fracture surface where cracks of the object being analyzed are generated, developed, and finally destroyed, and the relationship between the observation images and the destruction mode is learned. The accuracy of the first observation image in ultimately reaching destruction is calculated for each of the destruction modes. For the first observation image obtained by the imaging unit of the microscopic area of ​​the analyte, the accuracy of the destruction mode is obtained for multiple first observation images with different shooting positions, and the shooting position on the side with higher accuracy is represented as the observation condition that can improve the accuracy of the destruction mode.

5. A fracture analysis method, characterized in that: For a first observation image of the analyte obtained by the imaging unit of the microscopic region, the accuracy of the destruction mode is calculated, observation conditions including magnification and observation position that can improve the accuracy of the destruction mode are calculated, shooting conditions of the imaging unit of the microscopic region are adjusted, and the accuracy of the destruction mode and the observation conditions that can improve the accuracy of the destruction mode are displayed. Here, a learned neural network is used to process observation images including observation images of the fracture surface where a crack of the object to be analyzed is generated, developed, and finally destroyed, and the relationship between the observation images and the destruction mode is learned. The accuracy of the first observation image in ultimately reaching destruction is obtained for each of the destruction modes, and, The first observation image is divided into a plurality of regions, the accuracy of the failure mode is calculated for each of the divided regions, and the imaging position of the divided region with the highest accuracy is expressed as an observation condition that can improve the accuracy of the failure mode.

Images