TWI761892B - Defect detection mechanism and defect identification method for metal lamination manufacturing - Google Patents

Abstract

The invention provides a defect detection mechanism manufactured by metal lamination, which includes a platform, a fixture, a laser beam and metal powder injection assembly, a microphone (or a sound sensor) and an analysis computer. Wherein, the clamp is erected on the platform and used to fix the base material. The laser beam and metal powder injection assembly is erected above the platform and simultaneously projects the laser beam and the metal powder onto the base material for direct energy deposition. The microphone and the acoustic sensor are used to sense a set of sounds and acoustic signals generated during the melting and stacking process of the metal powder. The analysis computer receives and analyzes the set of sound and sound emission signals, and uses an identifier to identify the defects of the workpiece. Through the above-mentioned defect detection mechanism, the defects of the workpiece can be effectively detected.

Description

Defect detection mechanism and defect identification method for metal lamination manufacturing

The present invention relates to defect detection in metal lamination manufacturing, in particular to a defect detection mechanism and identification method in metal lamination manufacturing using direct energy deposition technology.

Additive Manufacturing (AM), also known as 3D printing technology, has been developed for many years, but for the additive manufacturing of metal materials (hereinafter referred to as metal additive manufacturing), there are still many technical problems to be overcome, such as the improvement of material properties. Due to the improvement of defects in the process, because of the huge commercial potential of metal laminate manufacturing, many advanced countries currently invest a lot of resources in actively researching related manufacturing equipment and process design.

At present, the mainstream technologies of metal lamination manufacturing can be roughly divided into powder bed fusion technology (Power Bed Fusion; PBF) and direct energy deposition technology (Directed Energy Deposition; DED). Parts of implements and thus have considerable potential. The principle of direct energy deposition technology is to transmit high-energy electromagnetic waves (usually using high-power lasers) and metal powder to a base material (such as a substrate) at the same time, and form a molten pool on the base material (molten pool), The metal powder is melted on the molten pool and combined and stacked to form a weld bead, which is finally formed into a workpiece. However, if the process conditions are not well set, such as insufficient laser power output, various defects such as spherical discontinuity defects (Balling), porosity (porosity) and other defects are easily generated on the structure of the workpiece. However, to the best of the applicant's knowledge, the existing methods for detecting defects mostly use optical inspection (for example, using CCD to capture images of the workpiece surface for analysis) or use process monitoring (monitoring the temperature of the material) to determine the structure of the workpiece. However, the above detection method is difficult to judge whether the internal structure of the workpiece is actually defective, so there is still room for improvement.

One of the objectives of the present invention is to improve the lack of the existing metal lamination detection technology, and further propose a brand-new defect detection mechanism and defect identification method, which can detect and identify the defects of the workpiece.

The reason is that, according to a defect detection mechanism for metal lamination manufacturing provided by the present invention, it is used to judge whether a workpiece has structural defects. The workpiece is formed on a base material, and the defect detection mechanism includes a platform and a fixture. , a laser beam and metal powder injection assembly, a microphone and an analysis computer. Wherein, the fixture is erected on the platform to fix the above-mentioned basic material. The laser beam and metal powder injection assembly is erected above the platform and directed to the base material. The laser beam and metal powder injection assembly simultaneously project the laser beam and the metal powder onto the base material, so that the metal powder is deposited by direct energy. way to melt and stack on the base material. The microphone is erected on the platform and pointed at the base material. The microphone senses a set of sound signals in the time domain generated during the melting and stacking process of the metal powder. The analysis computer is electrically connected to the microphone and receives the set of sound signals in the time domain. The analysis computer converts the set of sound signals in the time domain into a set of first frequency domain signals. The analysis computer also obtains a plurality of first frequency domain signals from the set of first frequency domain signals. For the first eigenvectors, the analyzing computer executes an identifier and inputs the first eigenvectors to the identifier to judge defects on the workpiece structure. Wherein, the platform may be driven to move horizontally relative to the laser beam and metal powder ejection assembly, or the laser beam and metal powder ejection assembly may be driven to move horizontally relative to the platform.

In this way, the user can perform direct energy deposition operations on the base material through the laser beam and the metal powder injection assembly, so that the metal powder can be melted and stacked on the base material, and the metal powder can be sensed by the microphone to melt and stack on the base material. The sound of the basic material process is analyzed by analyzing the computer and by executing an identifier to identify whether there is a structural defect in the formed workpiece.

The present invention further provides a defect identification method for metal lamination manufacturing, which is used to identify structural defects of a workpiece, the workpiece is made of metal powder and formed on a base material by direct energy deposition, The steps of the defect detection method include: step S1: obtaining a set of sound signals in the time domain generated during the process of forming the workpiece on the base material; step S2: converting the set of sound signals in the time domain into frequencies A set of first frequency domain signals on the domain, and a plurality of first feature vectors are obtained from the set of first frequency domain signals; Step S3: Input the first feature vectors to an identifier to identify defects of the workpiece.

The contents of the following embodiments are written according to Yang Weijun's research paper "Analysis of Sound and Acoustic Radio Signals for Spheroidization and Joint Defect Detection in Metal Lamination Manufacturing" (the instructor is Mr. Lu Mingquan), so the following embodiments disclose The scope should cover the entire contents of the above-mentioned dissertation, which will be explained together first.

In order to illustrate the technical features of the present invention in detail, the following examples are given and described in conjunction with the drawings as follows, wherein:

As shown in FIG. 1, a defect detection mechanism 1 for metal lamination manufacturing provided by an embodiment of the present invention is used to judge whether a workpiece W has a structural defect, and the workpiece W is formed on a base material 21 (this embodiment For example, on a substrate, the defect detection mechanism 1 includes a platform 10 , a fixture 20 , a laser beam and metal powder injection assembly 30 , a microphone 40 , a sound sensor 50 and an analysis computer 60 .

The platform 10 in this embodiment is a two-axis motion platform, which can be driven to move horizontally.

The fixture 20 is erected on the platform 10 to fix a base material 21 . In this embodiment, the base material 21 is a substrate, which is used to allow the metal powder P to be melted and stacked on the substrate.

The laser beam and metal powder injection assembly 30 is fixedly erected above the platform 10 and directed to the base material 21 . The laser beam and metal powder injection assembly 30 includes a nozzle 31 , and the structure of the nozzle 31 can be referred to as shown in FIG. 2 . The structure of the nozzle 31 includes a central opening 32, a central channel 33 and two outer channels 34, the inner channel and the outer channel 34 are both connected to the central opening 32, and the laser beam L of the high-power laser passes through the central channel 33 and the center The opening 32 is focused on the surface of the base material 21 to form a molten pool on the surface of the base material 21 , and the output power of the high-power laser is adjustable, such as 20 watts, 40 watts or 200 watts. The metal powder P and the inert gas flow are transmitted in the outer channel 34 and sprayed to the molten pool through the central opening 32. The metal powder P absorbs the energy of the laser beam L and melts and is stacked and arranged on the base material 21. The beam L and the metal powder P are irradiated onto the molten pool of the base material 21 , so that the metal powder P can be melted and stacked on the base material 21 by means of direct energy deposition to form the workpiece W.

It should be noted that, although in this embodiment, the laser beam and the metal powder ejection assembly 30 use the same nozzle 31 to emit the laser beam L and the metal powder P, in some cases, they may be separated There are two nozzles 31 arranged in the manner of the above, one of which emits the laser beam L, and the other emits the metal powder P, so the present embodiment should not be limited. In addition, although in this embodiment, the platform 10 can be driven to move horizontally relative to the laser beam and the metal powder ejection assembly 30, in some cases, the laser beam and the metal powder can also be ejected The assembly 30 may be driven to move horizontally relative to the platform 10, or both may be driven to move.

Please go back to Figure 1. The microphone 40 is erected above the platform 10 and points to the base material 21 . In this embodiment, the microphone 40 is a micro-electromechanical microphone produced by Knowles company, the model is SPM0408LE5H-TB, its sensitivity is -18dB±3dB, and the effective bandwidth is It is 100Hz~10kHz and has directivity. The microphone 40 is mainly used to sense a set of sound signals in the time domain generated by the thermal deformation and oscillation of the metal powder P during the melting and stacking process of the metal powder P.

The acoustic radiation sensor 50 is fixed on the fixture 20 by a torque wrench and is about 5 cm away from the position of the molten pool. In this embodiment, the acoustic radiation sensor 50 is the acoustic radiation sensor 50 produced by Kistler Company, the model is 8152C003032, the sensitivity is 57dB±10dB, and the effective frequency bandwidth is 50Hz~400kHz. The acoustic radiation sensor 50 is mainly used to sense a set of acoustic radiation signals in the time domain generated by the thermal deformation and oscillation of the metal powder P during the melting and stacking process, which is a solid wave signal. It should be noted that, in some cases, the acoustic radiation sensor 50 may also be erected on the base material 21 (substrate), which is not limited to this embodiment.

The analysis computer 60 electrically connects the microphone 40 and the acoustic radiation sensor 50 through a signal capture card 61 , and captures the sound signal and the acoustic radiation signal in the time domain sensed by the microphone 40 and the acoustic radiation sensor 50 . The signal capture card 61 is a signal capture card 61 produced by National Instruments and the model is PXle-6132. The analysis computer 60 is internally built with a program written by the program software LabVIEW to control the signal capture card 61 to capture the sound signal and the sound-radiation signal, and perform analog-to-digital conversion of the signal. In addition, the analysis computer 60 is also provided with an analysis program written in Python, so as to perform signal analysis on the captured sound signal and sound emission signal.

In terms of signal analysis, the analysis computer 60 performs fast Fourier transformation or wavelet packet transformation through the analysis program of Python to convert the set of sound signals in the time domain into a set of first frequency domain signals and respectively convert the set of sound signals in the time domain into a set of first frequency domain signals. The radio signal is converted into a set of second frequency domain signals, and then the analysis computer 60 obtains a plurality of first eigenvectors and second eigenvectors from the above-mentioned set of first frequency domain signals and the above-mentioned set of second frequency domain signals, and the analysis computer 60 Then execute an identifier and input the first frequency domain signals and the second frequency domain signals to the identifier, so as to identify whether the workpiece W has structural defects and identify the types of defects.

The construction of the identifier is described as follows: two different workpieces W are fabricated by direct energy deposition using the above-mentioned laser beam and metal powder ejection assembly 30, wherein one workpiece W has no defects and the other workpiece W has defects (such as joints). Defects (bonding defects), and the microphone 40 and/or the acoustic radiation sensor 50 sense a set of acoustic signals in the time domain and/or acoustic radiation in the time domain generated during the melting and stacking process of the metal powder P Signal. Afterwards, utilize the analysis computer 60 to convert the above-mentioned sound signal and/or the sound emission signal into a signal in the frequency domain, and obtain therefrom a plurality of eigenvectors corresponding to the defect-free workpiece W and a plurality of eigenvectors corresponding to the workpiece W with bonding defects multiple eigenvectors, and use these multiple eigenvectors to build a hidden Markov model identifier. The details of how the identifier is established can be found in the following paper: "Feedback Hidden Markov Model for Nickel-Based Material Cutting Tools "Development of Multiple Sensor Detection System for Wear and Wear", the author of the paper is Wang Chongying, and the instructor is Mr. Lu Mingquan. It should be noted that other types of models, such as neural network-like models, can also be used to build the identifier.

In the actual captured sound signal and acoustic signal, it may come from a variety of signal sources. As far as the sound signal is concerned, the source may come from the background sound signal of the machine, the sound signal that only the metal powder P hits the base material 21, the sound signal generated by the laser beam L forming a molten pool on the base material 21, the metal powder P only. The sound signal generated by the powder P has been melted before reaching the base material 21, the sound signal generated by the metal powder P reaching the base material 21 and being melted and stacked on the base material 21, wherein, the applicant has conducted a series of experiments and obtained from According to the experimental results, the source of the sound signal is mainly the sound generated when the metal powder P is melted and stacked on the base material 21 . The source of the acoustic signal is different. When the metal powder P simply hits the base material 21, there will be an obvious acoustic signal, and the acoustic signal is not affected by environmental noise.

The applicant has carried out the experiment of single-track direct energy deposition (that is, forming a metal line on the base material 21), and when the acoustic signals of normal single-track deposition and single-track deposition bonding defects (as shown in FIG. 3A and FIG. 4A ) are subjected to wavelet packet transformation The applicant has detected that the fluctuation of the acoustic signal generated by the normal monorail deposition in the frequency band of 312.5kHz~375kHz and the acoustic signal in the frequency range of 375kHz~437.5kHz (as shown in Figure 3B and Figure 3C) is greater than that of the monorail deposition joint defect. The degree of fluctuation of the radio signal (as shown in Figure 4B and Figure 4C). As for the sound signal (as shown in Fig. 5A and Fig. 6A ), the applicant has detected that the sound signal generated by normal single-track deposition in the frequency range of 0-6.25 kHz (as shown in Fig. 5B ) fluctuates more than the sound when the single-track deposition is defective The degree of fluctuation of the signal (as shown in Figure 6B).

In addition, for the spheroidization defect, the location of the spheroidization defect can be identified only by the sound signal in the time series (as shown in Figure 7A and Figure 8A ), and the sound signal generated by the normal single-track deposition is in the high-energy frequency region and the low-energy region. The spectral distribution in the frequency domain (as shown in FIGS. 7B and 7C ) is different from the spectral distribution of the sound signal of the single-track spheroidization defect in the high-energy region frequency region and the low-energy region frequency region (as shown in FIGS. 8B and 8C ). Acoustic signals do not see features. In this way, based on the analysis data based on the difference between the acoustic emission signal and the first and second frequency domain signals converted from the acoustic signal, the established identifier can be used to identify different types of defects.

It should be noted that the defects of direct energy deposition on the structure at least include bonding defects, balling defects, porosity and other types, although this embodiment only uses bonding defects and spheroidization. Defects are exemplified, but the detection mechanism 1 and the detection method in this embodiment can also be applied to other types of defect detection. In addition, this embodiment uses both the sound signal and the sound emission signal for detection, but it should be understood that it should be possible to use only the sound signal or the sound emission signal to detect defects, but the degree of recognition is only Maybe just a little bit worse.

Another embodiment of the present invention is to describe a defect identification method for metal lamination manufacturing, which is used to identify structural defects of a workpiece W, which is made of metal powder P and is formed by direct energy deposition. Formed on the base material 21, the steps of the defect detection method include:

Step S1 : obtaining a set of sound signals in the time domain generated during the process of forming the workpiece W on the base material 21 through the microphone 40 . A set of acoustic radiation signals in the time domain generated in the process of forming the workpiece W on the base material 21 are obtained by the acoustic radiation sensor 50 .

Step S2: Convert the group of sound signals in the time domain into a group of first frequency-domain signals in the frequency domain through fast Fourier transform or wavelet packet transformation, and convert the group of sound-radiation signals in the time domain into the frequency domain. A set of second frequency domain signals, and a plurality of first eigenvectors and a plurality of second eigenvectors are obtained from the set of first frequency domain signals and the set of second frequency domain signals.

Step S3: Input the first feature vectors and the second feature vectors to the identifier to identify the structural defects of the workpiece W.

Finally, it must be reiterated that the methods and constituent elements disclosed in the foregoing embodiments of the present invention are merely illustrative, and are not intended to limit the scope of the invention. Or replacement with other equivalent elements should still fall within the scope covered by the scope of the patent application of the present invention.

1: Testing agency 10: Platform 20: Fixtures 21: Basic Materials 30: Laser beam and metal powder injection assembly 31: Nozzle 32: Central opening 33: Central channel 34: Outside channel 40: Microphone 50: Acoustic sensor 60: Analytical Computers 61:Signal capture card L: laser beam P: metal powder W: workpiece S1-S3: Steps

The detailed structure, features, and identification methods of the defect detection mechanism and defect identification method for metal lamination manufacturing will be described in the following embodiments. However, it should be understood that the embodiments and drawings described below are only examples. It should not be used to limit the scope of the patent application of the present invention, wherein:

FIG. 1 is a schematic structural diagram of a defect detection mechanism manufactured by a metal laminate according to an embodiment; FIG. 2 is a schematic diagram of a laser beam and a metal powder injection assembly according to an embodiment; FIG. 3A is a graph showing the relationship between acoustic emission signal and time of normal monorail deposition; FIG. 3B and FIG. 3C are diagrams of the relationship between acoustic emission signals and time in different frequency bands through wavelet packet transformation in FIG. 3A; FIG. 4A is a graph showing the relationship between acoustic radiation signal and time of a single track deposition bonding defect; FIG. 4B and FIG. 4C are diagrams of the relationship between acoustic emission signals and time in different frequency bands through wavelet packet transformation in FIG. 4A; FIG. 5A is a graph showing the relationship between the acoustic signal and time of normal single-track deposition; FIG. 5B is a graph of the relationship between the sound signal and time in the frequency range of 0 to 62.5 kHz through wavelet packet transformation in FIG. 5A; FIG. 6A is a graph showing the relationship between the acoustic signal and time of a single track deposition bonding defect; FIG. 6B is a graph of the relationship between the sound signal and time in the frequency range of 0 to 62.5 kHz through wavelet packet transformation in FIG. 6A; FIG. 7A is a graph showing the relationship between the acoustic signal and time of normal single-track deposition; FIG. 7B and FIG. 7C are diagrams of the relationship between the sound signal and time in different frequency bands through wavelet packet transformation in FIG. 7A; 8A is a graph showing the relationship between the acoustic signal and time of the monorail spheroidization defect; FIG. 8B and FIG. 8C are graphs of the relationship between the sound signal in different frequency bands and time under the wavelet packet transformation of FIG. 8A; and FIG. 9 is a flow chart of the steps of the identification method of the embodiment.

1: Testing agency 10: Platform 20: Fixtures 21: Basic Materials 30: Laser beam and metal powder injection assembly 40: Microphone 50: Acoustic sensor 60: Analytical Computers 61:Signal capture card W: workpiece

Claims (6)

A defect detection mechanism manufactured by metal lamination is used for judging whether a workpiece has structural defects, the workpiece is formed on a base material, and the defect detection mechanism comprises: a platform; a fixture, erected on the platform and fixing the base material; A laser beam and metal powder injection assembly is erected above the platform and directed to the base material. The laser beam and metal powder injection assembly simultaneously projects the laser beam and the metal powder onto the base material, so that the metal powder is melted and stacked on the base material by means of direct energy deposition to form the workpiece; a microphone, erected on the platform and pointing at the base material, the microphone senses a set of sound signals in the time domain generated during the melting and stacking process of the metal powder; An analysis computer is electrically connected to the microphone and receives the set of sound signals in the time domain, the analysis computer converts the set of sound signals in the time domain into a set of first frequency domain signals, and the analysis computer extracts the set of first frequency domain signals from the set of sound signals and obtaining a plurality of first feature vectors, the analysis computer executes an identifier and inputs the first feature vectors to the identifier to identify the defects of the workpiece; Wherein, the platform can be driven to move horizontally relative to the laser beam and metal powder ejection assembly, or the laser beam and metal powder ejection assembly can be driven to move horizontally relative to the platform. The defect detection mechanism for metal lamination manufacturing as claimed in claim 1, further comprising an acoustic radiation sensor, the acoustic radiation sensor is erected on the fixture or the base material, and the acoustic radiation sensor receives the metal powder A set of acoustic radiation signals in the time domain generated during the melting and stacking process, the analysis computer is electrically connected to the acoustic radiation sensor and receives the set of acoustic radiation signals in the time domain, and the analysis computer converts the set of time domain acoustic signals The acoustic radiation signals on the above form a set of second frequency domain signals, the analysis computer obtains a plurality of second eigenvectors from the set of second frequency domain signals, and the analysis computer inputs the first eigenvectors and the second eigenvectors to identify the defects of the workpiece. A defect detection mechanism manufactured by metal lamination is used for judging whether a workpiece has structural defects, the workpiece is formed on a base material, and the defect detection mechanism comprises: a platform; a fixture, erected on the platform and fixing the base material; A laser beam and metal powder injection assembly is erected above the platform and directed to the base material. The laser beam and metal powder injection assembly simultaneously projects the laser beam and the metal powder onto the base material, so that the metal powder is melted and stacked on the base material by means of direct energy deposition to form the workpiece; an acoustic radiation sensor, which is erected on the fixture or the base material, the acoustic radiation sensor senses a set of acoustic radiation signals in the time domain generated during the melting and stacking process of the metal powder; an analysis computer, electrically connected to the acoustic emission sensor and receiving the acoustic emission signals in the time domain, the analysis computer converts the acoustic emission signals in the time domain into a second frequency domain signal, the analysis computer from the acoustic emission signal a set of second frequency domain signals to obtain a plurality of second feature vectors, the analysis computer executes an identifier and inputs the second feature vectors to the identifier to identify the defects of the workpiece; Wherein, the platform can be driven to move horizontally relative to the laser beam and metal powder ejection assembly, or the laser beam and metal powder ejection assembly can be driven to move horizontally relative to the platform. A defect identification method for metal lamination manufacturing is used to identify structural defects of a workpiece, the workpiece is made of metal powder and formed on a base material by direct energy deposition, and the defect detection method The steps include: S1: Obtain a set of sound signals in the time domain generated during the process of forming the workpiece on the base material; S2: Convert the set of sound signals in the time domain into a set of first frequency domain signals in the frequency domain, and obtain a plurality of first feature vectors from the set of first frequency domain signals; S3: Input the first feature vectors to an identifier to identify defects of the workpiece. The defect identification method for metal lamination manufacturing as claimed in claim 4, wherein in step S1, a set of acoustic signals in the time domain generated during the process of forming the workpiece on the base material are also obtained; in step S2 , also convert the set of acoustic radiation signals in the time domain into a set of second frequency domain signals in the frequency domain, and obtain a plurality of second eigenvectors from the set of second frequency domain signals; and in step S3, It is to input the first feature vectors and the second feature vectors to the identifier to identify the defects of the workpiece. A defect identification method for metal lamination manufacturing is used to identify structural defects of a workpiece, the workpiece is made of metal powder and formed on a base material by direct energy deposition, and the defect detection method The steps include: S1: Obtain a set of acoustic radiation signals in the time domain generated during the process of forming the workpiece on the base material; S2: Convert the set of acoustic radiation signals in the time domain into a set of second frequency domain signals in the frequency domain, and obtain a plurality of second eigenvectors from the set of second frequency domain signals; S3: Input the second feature vectors to an identifier to identify defects of the workpiece.

Images