KR-20260065290-A - Laser processing monitoring method and device using artificial intelligence with multi-layered structure

KR20260065290AKR 20260065290 AKR20260065290 AKR 20260065290AKR-20260065290-A

Abstract

A laser processing monitoring device using artificial intelligence having a multilayer structure according to the present invention comprises: microphone devices that acquire audio of different frequency bands from a laser processing surface and output audio signals accordingly; camera devices that capture images of the laser processing surface in different wavelength bands and output image information accordingly; and a control device comprising a laser processing monitoring module, the control device comprising: first lower AI functional units provided corresponding to each of the microphone devices, which detect multiple feature information from audio signals output by each of the microphone devices and generate lower-level laser processing status information by AI processing the detected multiple feature information; second lower AI functional units provided corresponding to each of the camera devices, which detect multiple feature information from video signals output by each of the camera devices and generate lower-level laser processing status information by AI processing the detected multiple feature information; and an AI functional unit that receives two or more of the lower-level laser processing status information output by the first lower AI functional units and the second lower AI functional units as feature information and generates final laser processing status information by AI processing.

Inventors

최해운

Assignees

계명대학교 산학협력단

Dates

Publication Date: 20260508
Application Date: 20241101

Claims (12)

Microphone devices that acquire audio of different frequency bands from a laser-processed surface and output an audio signal accordingly; Camera devices that capture images of the above-mentioned laser-processed surface in different wavelength bands and output corresponding image information; Each of the above microphone devices is provided in correspondence with each of them, First sub-AI functional units that detect multiple feature information from audio signals output by each of the above-mentioned microphone devices, and generate lower-level laser processing status information by AI processing the detected multiple feature information, Each of the above camera devices is provided in correspondence with each of them, Second sub-AI functional units that detect multiple feature information from video signals output by each of the above camera devices, and generate lower-level laser processing status information by AI processing the detected multiple feature information, A laser processing status monitoring device using multilayer artificial intelligence, comprising: a control device having a laser processing monitoring module composed of an AI function unit that receives two or more of the lower-level laser processing status information output by the first sub-AI function units and the second sub-AI function units as feature information, processes them via AI, and generates final laser processing status information.
In paragraph 1, The above AI function unit is, Intermediate AI functional units that generate laser processing state information by receiving two or more of the lower-level laser processing state information output by the above first lower AI functional units and the above second lower AI functional units as feature information and performing AI processing, Higher-level AI functional units that receive two or more of the intermediate-level laser processing status information output by the above-mentioned intermediate AI functional units as feature information and perform AI processing to generate higher-level laser processing status information, and A laser processing status monitoring device using multilayer artificial intelligence, characterized by being composed of a final AI function unit that generates final laser processing status information by combining upper-level laser processing status information output by the upper AI function units mentioned above.
In paragraph 2, User interface device; and When laser processing method information is input through the above user interface device, A first combination information for selecting lower-level laser processing status information to be provided to the intermediate AI functional units among the lower-level laser processing status information output by the first lower AI functional units and the second lower AI functional units according to the above laser processing method information, The system further comprises a combination determination unit that generates second combination information by selecting the intermediate-level laser processing status information to be provided to the upper-level AI functional units among the intermediate-level laser processing status information, and provides it to the intermediate-level AI functional units and the upper-level AI functional units. The above intermediate AI functional units, according to the above first combination information, Select some of the lower-level laser processing status information output by the first lower-level AI functional units and the second lower-level AI functional units, receive the selected lower-level processing status information, and generate intermediate-level laser processing status information. The above-mentioned upper AI functional units, according to the above-mentioned second combination information, A laser processing status monitoring device using multilayer artificial intelligence, characterized by selecting some of the mid-level laser processing status information output by the above-mentioned upper AI functional units, and generating upper-level laser processing status information by receiving the selected mid-level processing status information.
In paragraph 2, User interface device; and When laser processing method information is input through the above user interface device, A first combination information for selecting lower-level laser processing status information to be provided to the middle AI function units among the lower-level laser processing status information output by the first lower AI function units and the second lower AI function units according to the laser processing method information, and a first reflection ratio information representing the reflection ratio for each of the selected lower-level laser processing status information. By generating a second combination information for selecting the intermediate-level laser processing status information to be provided to the upper AI functional units among the above intermediate-level laser processing status information, and a second reflection ratio information representing the reflection ratio for each of the selected intermediate-level laser processing status information, Further comprising a combination determination unit provided by the above intermediate AI functional units and the above upper AI functional units; The above intermediate AI functional units, according to the above first combination information and first reflection ratio information, Select some of the lower-level laser processing status information output by the first lower-level AI functional units and the second lower-level AI functional units, and generate middle-level laser processing status information by AI processing the selected lower-level processing status information by assigning weights according to the first reflection ratio information. The above-mentioned upper AI functional units, according to the above-mentioned second combination information and second reflection ratio information, A laser processing status monitoring device using multilayer artificial intelligence, characterized by selecting some of the mid-level laser processing status information output by the above-described upper AI functional units, and generating upper-level laser processing status information by AI processing the selected mid-level processing status information by assigning weights according to the above-described second reflection ratio information.
In paragraph 1, A laser processing status monitoring device using multilayer artificial intelligence, further comprising a preprocessing unit for preprocessing an audio signal or video signal provided to the above-described control device, wherein the preprocessing unit is composed of one or more combinations of a notch filter, a low-pass/high-pass filter, a band-pass filter, a Wiener filter, a frequency domain noise removal (spectral subtraction), a noise gate, deep learning-based noise removal (denoising) recognition, wavelet denoising, an average value filter, and a mean/moving average filter.
In paragraph 1, The above AI function unit is deep-learned through training data, and The above training data are, It generates by dividing the original data into predetermined first length units, and The starting points for splitting the above training data are determined as points obtained by splitting the above original data into predetermined second length units, and A laser processing status monitoring device using multilayer artificial intelligence, characterized in that the second length is determined as a micro-unit shorter than the first length.
A method for monitoring a laser processing state in a monitoring device comprising: microphone devices that acquire audio of different frequency bands from a laser processing surface and output an audio signal accordingly; camera devices that capture images of the laser processing surface in different wavelength bands and output image information accordingly; and a control device. The first sub-AI functional units of the laser processing status monitoring module using multi-layer artificial intelligence equipped in the above-mentioned control device, Each of the above microphone devices is provided in correspondence with each of them, A step of detecting multiple feature information from audio signals output by each of the above microphone devices, and generating lower-level laser processing status information by AI processing the detected multiple feature information; The second sub-AI functional units of the laser processing status monitoring module using multi-layer artificial intelligence equipped in the above-mentioned control device, Each of the above camera devices is provided in correspondence with each of them, A step of detecting multiple feature information from video signals output by each of the above camera devices, and generating lower-level laser processing status information by AI processing the detected multiple feature information; and The AI functional unit of the laser processing status monitoring module using multilayer artificial intelligence equipped in the above-mentioned control device, A laser processing state monitoring method using multilayer artificial intelligence, comprising the step of receiving two or more of the lower-level laser processing state information output by the first lower AI functional units and the second lower AI functional units as feature information, processing them with AI, and generating final laser processing state information.
In Paragraph 7, The above AI functional unit is composed of intermediate AI functional units, upper AI functional units, and a final decision AI functional unit, and The above intermediate AI functional units receive two or more of the lower-level laser processing status information output by the first lower AI functional units and the second lower AI functional units as feature information and perform AI processing to generate laser processing status information; The above-described upper AI functional units receive two or more of the intermediate-level laser processing status information output by the intermediate AI functional units as feature information and perform AI processing to generate upper-level laser processing status information; and A laser processing state monitoring method using multilayer artificial intelligence, characterized by further including the step of the final decision AI function unit generating final laser processing state information by combining upper-level laser processing state information output by the upper-level AI function units.
In paragraph 8, The above monitoring device further comprises a user interface device; and a combination determination unit, The above combination determination unit, When laser processing method information is input through the above user interface device, A first combination information for selecting lower-level laser processing status information to be provided to the intermediate AI functional units among the lower-level laser processing status information output by the first lower AI functional units and the second lower AI functional units according to the above laser processing method information, The method further comprises the step of generating a second combination of selected intermediate-level laser processing status information to be provided to the upper AI functional units among the intermediate-level laser processing status information, and providing it to the intermediate AI functional units and the upper AI functional units. The above intermediate AI functional units, according to the above first combination information, Select some of the lower-level laser processing status information output by the first lower-level AI functional units and the second lower-level AI functional units, receive the selected lower-level processing status information, and generate intermediate-level laser processing status information. The above-mentioned upper AI functional units, according to the above-mentioned second combination information, A laser processing status monitoring method using multilayer artificial intelligence, characterized by selecting some of the mid-level laser processing status information output by the above-described upper AI functional units, and generating upper-level laser processing status information by receiving the selected mid-level processing status information.
In paragraph 8, The above monitoring device further comprises a user interface device; and a combination determination unit, The above combination determination unit, When laser processing method information is input through the above user interface device, A first combination information for selecting lower-level laser processing status information to be provided to the middle AI function units among the lower-level laser processing status information output by the first lower AI function units and the second lower AI function units according to the laser processing method information, and a first reflection ratio information representing the reflection ratio for each of the selected lower-level laser processing status information. By generating a second combination information for selecting the intermediate-level laser processing status information to be provided to the upper AI functional units among the above intermediate-level laser processing status information, and a second reflection ratio information representing the reflection ratio for each of the selected intermediate-level laser processing status information, The step of providing the above intermediate AI functional units and the above upper AI functional units; further comprising, The aforementioned intermediate AI functional units, According to the above first combination information and first reflection ratio information, Select some of the lower-level laser processing status information output by the first lower-level AI functional units and the second lower-level AI functional units, and generate middle-level laser processing status information by AI processing the selected lower-level processing status information by assigning weights according to the first reflection ratio information. The above-mentioned upper AI functional units, According to the above second combination information and second reflection ratio information, A laser processing status monitoring method using multilayer artificial intelligence, characterized by selecting some of the mid-level laser processing status information output by the above-described upper AI functional units, and generating upper-level laser processing status information by processing the selected mid-level processing status information with weights assigned according to the above-described second reflection ratio information.
In Paragraph 7, The above AI function unit further includes a step of deep learning through training data; and The above training data are, It generates by dividing the original data into predetermined first length units, and The starting points for splitting the above training data are determined as points obtained by splitting the above original data into predetermined second length units, and A laser processing state monitoring method using multilayer artificial intelligence, characterized in that the second length is determined as a micro-unit shorter than the first length.
In Paragraph 7, A laser processing status monitoring method using multilayer artificial intelligence, characterized by further including a preprocessing step for preprocessing an audio signal or video signal provided to the control device.

Description

Laser processing monitoring method and device using artificial intelligence with multi-layered structure The present invention relates to a machining and welding monitoring technology, and more specifically, to a laser processing monitoring method and apparatus using artificial intelligence having a multilayer structure that can improve laser processing quality by increasing the reliability of monitoring results regarding the laser processing status, such as machining or welding using a laser. This project (result) is the result of the LINC 3.0-Industry-Academic Joint Technology Development Project, titled "Development of a Deep Learning-Based Cutting Tool Wear and Life Prediction System," which was conducted in 2024 with funding from the Ministry of Education and support from the National Research Foundation of Korea. When performing laser material processing, which processes materials using a high-power laser, particularly laser direct metal fabrication technology that creates three-dimensional products using functional materials (metals, alloys, ceramics, etc.) directly from CAD data based on laser surface modification (surface heat treatment, melting, alloying, etc.) and laser cladding technology, materials are generally processed under a constant laser output. As such, even when processing is performed under a constant laser output (even when all other processing conditions are the same), the thickness of the processing layer is significantly affected by the physical properties, surface condition, size, thickness, and shape of the workpiece. In particular, when laser processing a relatively large area, such as in laser surface modification and direct metal forming technology, heat accumulates during the laser processing process, causing the surface temperature of the substrate to gradually rise. Consequently, the absorption rate of laser energy on the surface of the substrate gradually increases, resulting in a thicker processing layer and a problem where the metallurgical microstructure of the processing layer becomes non-uniform. Accordingly, a technology for controlling laser output based on the surface temperature of a processing material has been proposed in the past. An example of such technology is Patent No. 10-2001-0081868, published by the Korean Intellectual Property Office under the title "Real-time control method and apparatus for laser output based on measurement of surface temperature of a processing material and its application." This discloses a real-time laser output control method comprising the steps of: irradiating a laser beam generated from a laser oscillator onto the surface of a base material through a beam transmission unit and a focusing lens unit; measuring the temperature of infrared (IR) generated from the surface of the base material reacting with the laser beam irradiated onto the surface of the base material using a non-contact temperature sensor; inputting an analog signal corresponding to the measured temperature value into a Proportional-Integrated-Derivative (PID) board unit; converting the analog signal input to the PID board unit into a digital signal and comparing it with an electrical signal corresponding to a pre-entered target (set) temperature to check for an error; determining a feedback value for the error and inputting the result and temperature value to a control unit; and controlling the output of the laser beam according to the value input to the control unit. This method allows the output reflecting the feedback from the PID board control to be irradiated onto a specimen in real time, while simultaneously ensuring that the temperature of the molten pool or heat treatment unit is maintained consistently and stably at a set value. In addition, there is Patent No. 10-2019-0126770 published at the Korean Intellectual Property Office under the title "Laser Processing Device." This discloses a laser processing device comprising a thermal radiation measuring unit that measures the intensity of thermal radiation of a workpiece irradiated with laser light, and a judgment unit that determines the state of a workpiece, using the workpiece as the workpiece, based on the intensity of thermal radiation measured by the thermal radiation measuring unit. As such, although various technologies for laser-based heat treatment or welding have been proposed in the past, there was still an urgent need for the development of technologies capable of improving the performance of laser-based heat treatment or welding. FIG. 1 is a configuration diagram of a laser processing monitoring device according to a preferred embodiment of the present invention. FIG. 2 is a configuration diagram of a laser processing monitoring module using artificial intelligence having the multilayer structure of FIG. 1. Figure 3 is a structural diagram of the multilayer artificial intelligence of Figure 2. FIG. 4 is a diagram schematically illustrating the process of generating training data according t