CN116237545A

CN116237545A - Multi-target real-time monitoring device and system for additive manufacturing

Info

Publication number: CN116237545A
Application number: CN202310131717.9A
Authority: CN
Inventors: 李辉; 李秀花; 张臣; 雷诚
Original assignee: Wuhan University WHU
Current assignee: Wuhan University WHU
Priority date: 2023-02-16
Filing date: 2023-02-16
Publication date: 2023-06-09

Abstract

The invention discloses a multi-target real-time monitoring device and a system for additive manufacturing, wherein the multi-target real-time monitoring device comprises a multi-target real-time monitoring device and a brain-like computing chip of an heterogeneous fusion framework, the monitoring device comprises an excitation laser generator, a compressed sensing camera, an infrared thermometer, an infrared thermal imager, a photodiode, a galvanometer and a beam-splitting light path component consisting of three spectroscopes, and excitation laser emitted by the excitation laser generator reaches the galvanometer through the beam-splitting light path component and is scanned on a detection object through the galvanometer to serve as an excitation signal; after receiving the excitation signal, the detection object generates reflected light, and the reflected light is divided into multiple paths of different wavelength light paths through the light splitting light path component, and the paths respectively reach the corresponding light sensors for corresponding monitoring. The invention can perform compression sampling, high-speed transmission and reconstruction on the image data in the additive manufacturing process, monitor the temperature of the molten pool in real time and extract image video characteristics, thereby realizing the cross-scale multi-target information real-time monitoring of the temperature of the molten pool, sputtering, smoke plume and the profile of the component layer.

Description

Multi-target real-time monitoring device and system for additive manufacturing

Technical Field

The invention belongs to the field of additive manufacturing, relates to an additive manufacturing molten pool monitoring technology, and particularly relates to a multi-target real-time monitoring device and system for additive manufacturing.

Background

In the laser powder bed melting additive manufacturing, the behaviors of variables such as molten pool morphology, temperature, sputtering, surface morphology and the like directly influence the performance of manufactured parts, and the quality detection of products is mainly carried out in an off-line mode at present, so that a large amount of material resources, financial resources and manpower investment are required, and the quality of the products under the optimized process cannot be completely ensured. On-line monitoring and feedback regulation of key variables through various monitoring devices such as a high-speed camera and a photodiode are main methods for solving metallurgical defects and regulating and controlling the shape. However, existing imaging monitoring approaches are difficult to continuously capture the course of changes in the microsecond timescale due to the speed limitations of the imaging device. Meanwhile, the conventional monitoring mode can generate mass data and increase data storage and processing time on the premise of realizing high precision, so that the requirement of real-time monitoring, regulation and control of the additive manufacturing process can not be met. Under the technology of limited hardware equipment, a key basis of modern signal processing is Shannon sampling theory, namely, the number of discrete samples required by a signal to be reconstructed without distortion requires that the sampling frequency of the discrete samples is greater than or equal to one time of the highest frequency of the signal. However, according to Shannon sampling theory, the monitored huge video image data signal is difficult to meet the real-time requirement in the additive manufacturing process. In addition, the metal additive manufacturing process involves the problems of multiple dimensions and multiple physical fields, including surface defects of a molten pool and temperature information, a system capable of comprehensively simultaneously monitoring and feeding back surface defects of the molten pool and temperature signals is deficient at present, and image processing algorithms such as signal processing and artificial intelligence in the online monitoring process also face the problems of high target moving speed, low transmission processing efficiency and the like of part geometric features.

In addition, the multiple parameter monitoring devices have complex structure and huge volume, and seriously affect the metal additive manufacturing light path and the powder feeding mechanism, so that development of a set of data compression sampling device and system is needed to be capable of efficiently and rapidly acquiring and reconstructing original multi-element signals with high quality so as to realize in-situ real-time monitoring of additive manufacturing, reduce the occupied space of the detection device and improve the flexibility of laser cladding and powder feeding.

Disclosure of Invention

Aiming at the difficult problems of harsh online visual monitoring conditions and low treatment efficiency in the laser powder bed fusion manufacturing process, the invention discloses a multi-target real-time monitoring device and system for additive manufacturing. The system aims at the difficult problems of small area of a molten pool, fast solidification, high temperature and multiple interference factors in online monitoring, and an infrared temperature measuring system and a photodiode are used for efficiently monitoring the temperature signal of the molten pool. The invention further aims at the problems of large data quantity, low transmission efficiency and low signal to noise ratio of monitoring signals in the laser powder bed melting and material increasing manufacturing process, designs and builds a high-speed imaging camera based on a compressed sensing technology, realizes real-time continuous capturing of the processing process, effectively removes redundant data in signals by utilizing a DMD mask matrix aiming at molten pool signals, does not influence the detection precision of a system, and reconstructs an original image at a decoding end through a random demodulation reconstruction technology so as to shorten the image transmission time, thereby realizing real-time monitoring of the processing process. The compressed sensing technology provides a solution for solving the problems of low resolution ratio and large data volume of effective monitoring signals in the laser powder bed melting additive manufacturing process. Meanwhile, in the monitoring process, the signal complex feature edge computing method is integrated into a brain-like computing chip of an heterogeneous fusion architecture, and the inherent memory access rate limit of the von neumann architecture is broken through by means of the memory integrated design of the brain-like computing chip, so that TB-level monitoring signal processing per minute can be completed. And finally, integrating a compression sensing camera, a temperature monitoring light path and a brain-like computing chip to realize real-time monitoring of the temperature of a molten pool, sputtering, smoke plume and cross-scale multi-target information of the profile of the component layer in the processing process.

The above object of the present invention is achieved by the following means:

a multi-target real-time monitoring device comprises

The galvanometer is used for scanning the excitation laser and collecting the reflected light;

the excitation laser generator is used for emitting excitation laser with set wavelength and is connected with the vibrating mirror through the beam splitting light path component;

the detection assembly is connected with the vibrating mirror through the light splitting optical path assembly and comprises a plurality of light sensors for detecting different parameters, and the working optical path is split according to the wavelength through the light splitting optical path assembly, so that the light sensors work in different wave bands without mutual interference;

excitation laser emitted by the excitation laser generator reaches the vibrating mirror through the beam splitting optical path component and is scanned on a detection object through the vibrating mirror to serve as an excitation signal; the detection object receives the excitation signal and then generates reflected light, the reflected light is divided into a plurality of paths of different wavelengths through the light splitting optical path component, and the paths of the reflected light respectively reach the corresponding light sensors, so that the multi-target real-time synchronous monitoring of the same object is realized.

Preferably, the three photo sensors are respectively a compressed sensing camera, an infrared thermometer and a photodiode.

Preferably, the beam splitting optical path component comprises three beam splitters which are arranged in parallel and semi-transparent and semi-reflective, the first beam splitter is a band-pass beam splitter, and the second beam splitter and the third beam splitter are long-pass beam splitters with sequentially increased working wavelengths or short-pass beam splitters with sequentially reduced working wavelengths;

the excitation laser generator and the vibrating mirror are respectively arranged at two opposite sides of the first spectroscope, and the bandpass wavelength of the first spectroscope is matched with the excitation laser; the three light sensors are sequentially arranged on the reflection light path of the second spectroscope, the reflection light path of the third spectroscope and the transmission light path of the third spectroscope; the working wavelengths of the three light sensors are sequentially increased or decreased according to the types of the second spectroscope and the third spectroscope, and are staggered with the wavelength of the excitation laser;

excitation laser emitted by the excitation laser generator reaches the vibrating mirror through the first spectroscope with the band-pass type, and an excitation signal is generated on a measurement object through scanning of the vibrating mirror; the detection object receives the excitation signal and then generates reflected signal light, the reflected signal light reaches the first spectroscope after being collected by the vibrating mirror, the reflected signal light is reflected by the first spectroscope and then is sent to the corresponding light sensor after being separated by the second spectroscope and the third spectroscope to obtain different length wavelengths.

Preferably, the working wavelengths of the compression sensing camera, the infrared thermometer and the photodiode are sequentially increased, and correspondingly, the second spectroscope and the third spectroscope are long-pass spectroscopes with sequentially increased working wavelengths, and the working wavelengths are respectively marked as A1 and A2; the compressed sensing camera is arranged on a reflection light path of the second spectroscope, the working wavelength is smaller than or equal to A1, the infrared thermometer is arranged on a reflection light path of the third spectroscope, the working wavelength is positioned between A1 and A2, the photodiode is arranged on a transmission light path of the third spectroscope, and the working wavelength is larger than A2.

Preferably, the multi-target real-time monitoring device further comprises a thermal infrared imager for synchronously aligning the detection objects.

Preferably, the excitation laser generator comprises a laser and a beam expander, wherein the beam expander is used for expanding laser emitted by the laser and then emitting the laser.

Preferably, the compressed sensing camera comprises a high-speed photoelectric detector, a femtosecond pulse laser, a digital micromirror array, a beam splitter and an acousto-optic deflector which are sequentially arranged on the same optical path; the high-speed photoelectric detector is arranged on a reflection light path of the digital micro-mirror array;

the femtosecond pulse laser emits femtosecond pulse laser which sequentially passes through the digital micro mirror array, the beam splitter and the acousto-optic deflector and then is used as a perception camera excitation signal, the femtosecond pulse laser sequentially passes through the second beam splitter and the first beam splitter to be reflected to the vibrating mirror, and the vibrating mirror scans on a detection object;

the detection object receives the excitation signal of the sensing camera to generate a reflection signal, the reflection signal is collected by the vibrating mirror, then reflected by the first spectroscope and the second spectroscope to reach the compressed sensing camera, and the reflection signal is detected by the high-speed photoelectric detector after sequentially passing through the acousto-optic deflector, the beam splitter and the digital micro-mirror array.

Preferably, a total reflection mirror for changing the direction of the light path and a lens for condensing are also arranged between the acousto-optic deflector and the second beam splitter of the compressed sensing camera.

The invention also provides a multi-target real-time monitoring system for additive manufacturing, which comprises any multi-target real-time monitoring device and a brain-like computing chip of heterogeneous fusion architecture, wherein a neural network algorithm is arranged in the brain-like computing chip and is used for extracting characteristics of information acquired by a compressed sensing camera.

Compared with the prior art, the invention has the following beneficial effects:

the compressed sensing camera designed by the invention can realize compressed sampling of the surface signal of the molten pool in the processing process. Redundant data in the signals are effectively removed by designing a mask matrix aiming at molten pool signals, the detection precision of the system is not affected, an image reconstruction algorithm is optimized to shorten the image transmission time, and real-time monitoring of the processing process is realized.

The designed molten pool temperature monitoring light path combines the advantages of an infrared temperature measuring system and a photodiode, and can acquire higher temperature information of the molten pool at a high speed;

according to the invention, the molten pool morphology signal compressed and sampled by the compressed sensing camera is subjected to high-speed and high-efficiency extraction of the molten pool surface characteristics and reconstruction of the original signal by using a brain-like chip (48TFLOPS@FP16) at the output end based on a neural network and a deep compressed sensing network, so that the algorithm running rate is greatly improved. The brain-like chip utilizes the physical characteristics of integration of edge calculation and memory, is matched with the edge calculation naturally, and finishes the pretreatment of the original data at the original data collection end, thereby reducing the generation of redundant data and reducing the cost of data storage, processing and transmission.

The invention establishes a dynamic visual, infrared thermal imager, diode and multi-metal additive manufacturing on-line monitoring system, acquires the dynamic behavior of a molten pool and the characteristic information of macro-micro tissues, and can monitor the molten pool, smoke plume and workpiece morphology simultaneously in multiple spatial scales. The compressed sensing edge computing method of the multi-source monitoring signal is provided, and is combined with the autonomous controllable brain-like computing chip for the first time, the compressed sensing monitoring signal is rapidly processed on line at the edge end, and rapid and efficient on-line detection is realized.

The multi-target real-time monitoring system built by the invention combines multiple monitoring technology light paths into one path, and adopts one vibrating mirror, so that the occupied space is greatly saved, the flexibility of additive manufacturing is greatly improved, and the increase of the manufacturing speed and the capability of manufacturing complex workpieces are facilitated.

Drawings

FIG. 1 is a schematic diagram of a multi-target real-time monitoring system for additive manufacturing.

Fig. 2 is a diagram of a temperature monitoring light path of the multi-target real-time monitoring device according to the present invention.

Fig. 3 is a diagram of the compressed sensing camera according to the present invention.

Fig. 4 shows a compressed sensing AI chip feature extraction flow diagram.

The device comprises a 1-photodiode, a 2-infrared thermometer, a 3-compression sensing camera, a 4-excitation laser generator, a 41-laser, a 42-beam expander, a 5-brain-like calculation chip, a 6-infrared thermal imager, a 7-first spectroscope, an 8-second spectroscope, a 9-third spectroscope, a 10-galvanometer, an 11-molten pool, a 12-laser wire column 1, a 13-laser wire column 2, a 14-lens, a 15-reflecting mirror, a 16-acousto-optic deflector, a 17-beam splitter, a 18-digital micro mirror array 19-femtosecond pulse laser, a 20-high-speed photoelectric detector, a 21-workpiece and a 22-working platform.

Detailed Description

Embodiments of the present invention are described in further detail below with reference to the accompanying drawings and examples. The following examples are illustrative of the invention but are not intended to limit the scope of the invention.

As shown in fig. 1 to 3, the present invention provides a multi-target real-time monitoring system for additive manufacturing, which comprises a multi-target real-time monitoring device and a brain-like computing chip 5 of heterogeneous fusion architecture;

the multi-target real-time monitoring device comprises a galvanometer 10, a beam splitting optical path component, an excitation laser generator 4 and a detection component, wherein the excitation laser generator 4 is used for emitting excitation laser with set wavelength and is connected with the galvanometer 10 through the beam splitting optical path component;

the detection component is also connected with the vibrating mirror 10 through a light splitting light path component and comprises a plurality of light sensors for detecting different parameters, and the working light path is split according to the wavelength through the light splitting light path component so that the plurality of light sensors work in different wave bands without mutual interference;

the excitation laser emitted by the excitation laser generator 4 reaches the galvanometer 10 through the beam splitting optical path component and is scanned on a detection object through the galvanometer 10 to serve as an excitation signal; the detection object receives the excitation signal and then generates reflected light, the reflected light is divided into a plurality of paths of different wavelengths through the light splitting optical path component, and the paths of the reflected light respectively reach the corresponding light sensors, so that the multi-target real-time synchronous monitoring of the same object is realized.

The brain-like computing chip 5 is internally provided with a neural network algorithm for extracting characteristics of information acquired by one or more light sensors.

The excitation laser generator 4 is utilized to emit excitation laser, and the beam splitting optical path component filters the excitation laser and then scans the excitation laser in a molten pool 11 and a nearby area of additive manufacturing through the vibrating mirror 10; and then, monitoring different spatial scales and target parameters of the molten pool 11 and the nearby areas by using different light sensors, and performing feature extraction on monitoring information obtained by one or more light sensors through the brain-like calculation chip 5 to obtain multi-target trans-scale information of the temperature information, sputtering, smoke plume and component layer profile of the molten pool 11.

As a specific embodiment, the photo sensors adopted in the invention are respectively a compression sensing camera 3, an infrared thermometer 2 and a photodiode 1; according to the invention, the infrared thermometer 2 and the photodiode 1 accurately measure the temperature information of the molten pool 11 at a high speed, the infrared thermometer 2 can monitor the temperature information of the molten pool 11 at 3000 ℃, the photodiode 1 can rapidly measure the temperature signal of the molten pool 11, and the advantages of the infrared thermometer 2 and the photodiode are complementary, so that the rapid acquisition and monitoring of the temperature characteristics of the higher molten pool 11 are realized.

The compressed sensing camera 3 module performs sparse sampling on the original signal through a dynamic mask generated by a digital micromirror array 18 (DMD), so that the transmission data quantity is greatly reduced, a random demodulation reconstruction technology is adopted at a decoding end, and the online high-precision reconstruction of the surface morphology of the molten pool 11 can be realized through a brain-like calculation chip 5 with a built-in neural network algorithm.

The brain-like computing chip 5 breaks through the inherent memory access rate limit of the von Neumann architecture by means of the memory-like computing chip memory integrated design, and the TB-level monitoring signal processing per minute is completed. And transmitting the system image feature extraction algorithm and the compressed sensing image reconstruction algorithm to the brain-like calculation chip 5. And the processing efficiency of the monitoring signal is improved through the cooperative optimization of software and hardware, so that the on-line identification of various abnormal manufacturing state characteristics is realized.

As a preferred embodiment, as shown in fig. 2, the beam splitting optical path assembly includes three beam splitters (dichroic mirrors) that are disposed parallel to each other and semi-transparent and semi-reflective, the first beam splitter 7 is a band-pass beam splitter, the second beam splitter 8 and the third beam splitter 9 are long-pass beam splitters with sequentially increased working wavelengths or short-pass beam splitters with sequentially decreased working wavelengths;

the excitation laser generator 4 and the galvanometer 10 are respectively arranged at two opposite sides of the first spectroscope 7, and the bandpass wavelength of the first spectroscope 7 is matched with the excitation laser; the three light sensors are respectively arranged on the reflection light path of the second spectroscope 8, the reflection light path of the third spectroscope 9 and the transmission light path of the third spectroscope 9; the working wavelengths of the three light sensors are sequentially increased or decreased according to the types of the second spectroscope 8 and the third spectroscope 9 so as to realize light path signal distinction; the wavelength of the excitation laser is staggered, so that signal interference is prevented;

for example, the second spectroscope 8 and the third spectroscope 9 are long-pass spectroscopes with sequentially increased working wavelengths, and the working wavelengths of the corresponding three light sensors are sequentially increased according to the types of the second spectroscope 8 and the third spectroscope 9;

the second spectroscope 8 and the third spectroscope 9 are short-pass spectroscopes with the working wavelengths sequentially reduced; the operating wavelengths of the corresponding three light sensors are sequentially reduced according to the types of the second spectroscope 8 and the third spectroscope 9.

The signal monitoring principle of the invention is as follows:

the excitation laser emitted by the excitation laser generator 4 reaches the galvanometer 10 through the first spectroscope 7 with a band-pass type, and an excitation signal is generated on a measurement object by scanning through the galvanometer 10; the detection object receives the excitation signal and then generates reflected signal light, the reflected signal light reaches the first spectroscope 7 after being collected by the vibrating mirror 10, the reflected signal light is reflected by the first spectroscope 7 and then is separated into wavelengths with different lengths by the second spectroscope 9 and the third spectroscope 9, and the wavelengths are sent to the corresponding light sensor.

By the design of the light splitting light path component, three light sensors (the compression sensing camera 3, the infrared thermometer 2 and the photodiode 1) can share one path of reflected light signal, so that light path component equipment is greatly reduced, and most importantly, synchronous monitoring is truly realized by the three light sensors, time difference is eliminated in technical principle, and multi-target information extracted by the invention is monitoring information which is completely overlapped in time and space.

As a preferred embodiment, the exciting laser generator 4 includes a laser 41 and a beam expander 42 for increasing the diameter of the laser spot, the laser 41 is a pulse laser, and the beam expander 42 can ensure the coverage of the exciting laser, so that the whole molten pool 11 and the vicinity thereof are sufficiently covered.

As a preferred embodiment, the invention additionally designs an infrared thermal imager which is synchronously aligned with the area of the molten pool 11, and the infrared thermal imager has the advantage of wide measurement range and can acquire temperature distribution information, so that the infrared thermal imager is effectively complemented with the monitoring of the material adding process; since the infrared thermal imager itself is a wide area sensor, it is not necessary to add the infrared thermal imager to the spectroscopic assembly of the present invention, and the temperature measurement is performed separately for the molten pool 11 and the vicinity.

As a preferred embodiment, as shown in fig. 3, the compression sensing camera 3 includes a high-speed photodetector 20, and a femtosecond pulse laser 19, a digital micromirror array 18, a beam splitter 17, and an acousto-optic deflector 16 sequentially disposed on the same optical path; the high-speed photodetector 20 is disposed on the reflected light path of the digital micromirror array 18; the femtosecond pulse laser 19 emits femtosecond pulse laser, which sequentially passes through the digital micro-mirror array 18, the beam splitter 17 and the acousto-optic deflector 16 and then is used as a sensing camera excitation signal, the sensing camera excitation signal sequentially passes through the second beam splitter 8 and the first beam splitter 7 and is reflected to the vibrating mirror 10, the sensing camera excitation signal is scanned on a detection object through the vibrating mirror 10, the detection object receives the sensing camera excitation signal to generate a reflection signal, the reflection signal is collected by the vibrating mirror 10 and then is reflected by the first beam splitter 7 and the second beam splitter 8 and then reaches the compressed sensing camera 3, and the reflection signal sequentially passes through the acousto-optic deflector 16, the beam splitter 17 and the digital micro-mirror array 18 and then is detected by the high-speed photoelectric detector 20.

The detection light used by the compressed sensing camera 3 is ultra-short (femtosecond) light pulse, and a mask image of random coding is quickly generated in a time domain by combining a space-time-frequency dispersion technology, so that the limitation of the speed of the existing coding device is broken through, and the ultra-high coding speed is realized. The high-speed continuous sampling capability of the high-speed photoelectric detector 20 is combined, so that the dynamic behavior of the molten pool 11 is monitored at high speed on line. When the monitoring data is transmitted to the signal processing end at high speed, the limitation of the Nyquist sampling law is broken through according to a random demodulation reconstruction technology, and high-precision reconstruction of the monitoring image video information is realized.

Still preferably, a total reflection mirror 15 for changing the direction of the optical path and a lens 14 for condensing the femtosecond pulse laser are further provided between the acousto-optic deflector 16 and the second beam splitter 8 of the compressed sensing camera 3 of the present invention, and the femtosecond pulse laser can be condensed by the lens 14.

It should be noted that, the basic technical problem to be solved by the present invention is how to realize simultaneous monitoring of multiple parameters, so the above-mentioned compressed sensing camera 3 is only a preferred real-time mode, and does not represent the limitation of the protection scope of the present invention, and other compressed sensing cameras or optical sensors can be adopted to replace the compressed sensing camera, so that the implementation of the technical scheme of the present invention is not affected.

Based on the design, the invention provides a specific multi-target real-time monitoring system parameter design case, which comprises the following specific steps:

an infrared temperature measuring system is formed by an infrared thermometer 2 (the spectrum range is 0.8-1.0 μm) and an infrared thermal imager 6 (the spectrum range is 0.5-0.54 μm and 0.8 μm);

the first spectroscope 7 is a spectroscope with a 1.0-1.1 mu m band-pass, the second spectroscope 8 is a spectroscope with a 0.81 mu m long pass, and the third spectroscope 9 is a spectroscope with a 1.1 mu m long pass; the working wavelength of the laser 41 is selected to be 1.0-1.1 μm, and 1064nm is specifically selected; the working spectrum range of the infrared thermometer 2 is 0.81-1.0 mu m; the working spectrum range of the thermal infrared imager 6 is as follows: 0.5-0.54 μm and 0.8 μm; the working spectrum range of the compressed sensing camera 3 is 0.78-0.8 mu m, and particularly, 0.8 mu m can be selected; the working spectrum range of the photodiode 1 is 1.1 μm to 3.0 μm;

the working wave band of the vibrating mirror 10 is 780nm, and the working wavelength of the femtosecond pulse laser 19 is 0.8 mu m;

the brain-like computing chip 5 adopts the model 48TFLOPS@FP16, an embedded image feature extraction algorithm and a compressed sensing algorithm (both are known in the prior art), the brain-like chip is used for improving the system image feature extraction rate, and the algorithm is fused into the brain-like chip at a signal output end through the image feature extraction algorithm and the compressed sensing algorithm to reconstruct the original signal rapidly, so that the surface features of the molten pool 11 are extracted and reconstructed at high speed and high efficiency.

Specifically, a laser 41 (wavelength 1064 nm) emits a laser beam, the laser beam passes through a beam expander 42 to expand the spot diameter and then passes through a first spectroscope 7 (1.0-1.1 μm bandpass) and a galvanometer 10 to a molten pool 11 as excitation laser; of the reflected light rays of the molten pool 11, light rays in a spectral range of 0.5-0.54 mu m and 0.8 mu m are received by the thermal infrared imager 6; the reflected light rays in the rest spectral ranges reach the first spectroscope 7 after being deflected by the vibrating mirror 10, and the reflected light rays with the wavelengths of 1.0-1.1 μm are reflected to the second spectroscope 8; among the reflected light rays reaching the second spectroscope 8, the light rays with the wavelength of less than 0.81 mu m are totally reflected to the compressed sensing camera 3 and are used by the compressed sensing camera 3, and the reflected light rays with the wavelength of more than 0.81 mu m are totally transmitted to the third spectroscope 9; among the reflected light rays reaching the third spectroscope 9, the light rays with the wavelength of less than 1.1 μm are totally reflected to the infrared thermometer 2, are used by the operation of the infrared thermometer 2 (0.81 μm to 1.0 μm), and the reflected light rays with the wavelength of more than 1.1 μm are totally transmitted to the photodiode 1, and are used by the operation of the photodiode 1 (1.1 μm to 3.0 μm).

For the compressed sensing camera 3, the operating wavelength of the femtosecond pulse laser 19 is 0.8 μm; the femtosecond pulse laser 19 outputs 800nm laser beam, passes through the DMD, the beam splitter 17, the acousto-optic deflector 16, the reflecting mirror 15 and the lens 14, the second beam splitter 8, the first beam splitter 7 and the galvanometer 10 act to reach the surface of the molten pool 11, and the beam (laser line column 2) reflected by the molten pool 11 passes through the galvanometer 10, the first beam splitter 7, the lens 14, the reflecting mirror 15, the acousto-optic deflector 16, the beam splitter 17 and the DMD and finally forms an image on the high-speed photoelectric detector 20. The femtosecond pulse laser 19 has the advantages of high light output power, stable wavelength, spectrum width up to 80nm, high isolation, ultra-stable laser, tunable middle infrared and the like, and can be suitable for rapid monitoring of different temperature environments. The vibrating mirror 10 operates in a wavelength band of 780nm for adjusting the beam angle to machine and scan the workpiece 21. To match sensors of different wavelength ranges, the second beam splitter 8 is selected to be 810nm long pass for selecting (reflecting and projecting) light; the lens 14 is a conventional lens with high light transmittance and working wavelength of 780nm and is used for converging light; the reflector 15 is a conventional lens with high light transmittance and working wavelength of 780nm and is used for reflecting light rays; the acousto-optic deflector 16 selects AODF4085 of G & H, the direction of laser is controlled by a circuit signal, the response delay is shortened, and the beam angle can be adjusted to realize the scanning of the workpiece 21. The beam splitter 17 operates in a wavelength band of 780nm, and the reflected light (workpiece 21 information) is received by the detector by separating the round-trip light; the DMD selects V-650L of the vilux, and the monitored light signals are selectively reflected through a mask, so that the image signal compression and decoding of the acquisition end are realized.

The operating wavelength of the photo sensor and the operating wavelength of the spectroscope are not limited to the above-mentioned example parameters, and the operating wavelength of the photo sensor or the spectroscope may be adjusted according to the actual situation as long as the reflected light can be split into the corresponding photo sensor according to the wavelength and the reflected light does not interfere with each other.

The above embodiments are only for illustrating the present invention, and are not limiting of the present invention. While the invention has been described in detail with reference to the embodiments, those skilled in the art will appreciate that various combinations, modifications, and substitutions can be made thereto without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A multi-target real-time monitoring device is characterized by comprising

2. The multi-target real-time monitoring device of claim 1, wherein: the light splitting optical path component comprises three semi-transparent semi-reflective spectroscopes which are arranged in parallel, wherein the first spectroscope is a band-pass spectroscope, and the second spectroscope and the third spectroscope are long-pass spectroscopes with sequentially increased working wavelengths or short-pass spectroscopes with sequentially reduced working wavelengths;

3. The multi-target real-time monitoring device of claim 2, wherein: the three light sensors are respectively a compressed sensing camera, an infrared thermometer and a photodiode.

4. A multi-target real time monitoring device according to claim 3, wherein: the working wavelengths of the compression sensing camera, the infrared thermometer and the photodiode are sequentially increased, and correspondingly, the second spectroscope and the third spectroscope are long-pass spectroscopes with sequentially increased working wavelengths, and the working wavelengths are respectively marked as A1 and A2; the compressed sensing camera is arranged on a reflection light path of the second spectroscope, the working wavelength is smaller than or equal to A1, the infrared thermometer is arranged on a reflection light path of the third spectroscope, the working wavelength is positioned between A1 and A2, the photodiode is arranged on a transmission light path of the third spectroscope, and the working wavelength is larger than A2.

5. The multi-target real time monitoring device according to claim 4, wherein: a thermal infrared imager for synchronizing alignment of the inspection objects is also included.

6. The multi-target real time monitoring device of claim 5, wherein: the excitation laser generator comprises a laser and a beam expander, wherein the beam expander is used for expanding laser emitted by the laser and then emitting the laser.

7. The multi-target real time monitoring device according to claim 4, wherein: the compressed sensing camera comprises a high-speed photoelectric detector, a femtosecond pulse laser, a digital micromirror array, a beam splitter and an acousto-optic deflector which are sequentially arranged on the same optical path; the high-speed photoelectric detector is arranged on a reflection light path of the digital micro-mirror array;

8. The multi-target real time monitoring device of claim 7, wherein: and a total reflection mirror for changing the direction of the light path and a lens for condensing are also arranged between the acousto-optic deflector and the second beam splitter of the compressed sensing camera.

9. The multi-target real-time monitoring system for additive manufacturing is characterized by comprising the multi-target real-time monitoring device and a brain-like computing chip with a heterogeneous fusion architecture, wherein the brain-like computing chip is internally provided with a neural network algorithm and is used for extracting characteristics of information acquired by a compressed sensing camera.