An arc fuse additive porosity defect detection method based on multiple sensing signals

Abstract

A kind of arc fuse additive porosity defect detection method based on multiple sensing signals, first, build online monitoring platform, and online monitoring platform includes arc fuse mechanical arm, high-speed CCD camera, current voltage acquisition module, spectrum acquisition module;Then arc fuse additive process real-time collection set high-speed CCD image information, current voltage signal, spectrum signal, the size and position information of porosity defect are obtained by CT scanning to the printed piece, and the porosity defect feature set is analyzed and constructed;Porosity defect is divided into: no porosity defect, small porosity defect, medium and large porosity defect, and the high-speed CCD image, current voltage signal, spectrum signal collected at corresponding moment are corresponded, and the porosity defect feature data set is extracted;Finally, according to porosity defect feature data set, the directional optimization of arc fuse additive process is carried out;The present application realizes nondestructive, fast detection to porosity defect in arc fuse additive manufacturing.

Description

Technical Field

[0001] The present invention relates to the technical field of arc fuse additive manufacturing, and in particular to a method for detecting porosity defects in arc fuse additive manufacturing based on multiple sensor signals. Background Art

[0002] Arc-fuse additive manufacturing (AM) technology is widely used in aerospace, automotive, and other fields due to its high efficiency and flexibility. The physical mechanisms of the arc-material interaction during AM are extremely complex, which can lead to defects such as cracks, holes, and inclusions inside and outside parts. These defects can severely impact part performance and service life. Therefore, to ensure the reliability of AM technology and the quality of finished parts, implementing online monitoring and defect detection is crucial.

[0003] Online monitoring technology enables real-time monitoring of the manufacturing process, promptly identifying and warning of potential defects, thereby ensuring the quality and performance of manufactured parts. This technology enables early identification and timely intervention of defects during the manufacturing process, significantly improving product reliability and safety. This is crucial for advancing high-end equipment manufacturing and the development of cutting-edge technologies.

[0004] The patent application entitled "Metal Surface Defect Detection Method and System Based on Improved YOLOX" (Publication Number: CN117635521A) receives annotated metal defect detection samples, pre-processes the annotated metal defect detection samples, and obtains a metal defect dataset. However, similar methods of using datasets for training require manual labeling, which has problems such as a large amount of labeled data. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the present invention provides a multi-sensor signal-based porosity defect detection method for arc fuse additive manufacturing. During the arc fuse additive manufacturing process, high-speed CCD image information, current and voltage signals, and spectral signals are collected. The size and location of the porosity defects are determined by performing a CT scan on the printed part. A feature dataset corresponding to the defect is then extracted from the signal. Porosity defects in the arc fuse additive manufacturing process are then detected using machine learning, enabling real-time, non-destructive, and rapid detection of porosity defects in arc fuse additive manufacturing.

[0006] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0007] A method for detecting porosity defects in arc fuse additive materials based on multiple sensor signals comprises the following steps:

[0008] The first step is to build an online monitoring platform, which includes an arc fuse robot arm, a high-speed CCD camera, a current and voltage acquisition module, and a spectrum acquisition module.

[0009] In the second step, high-speed CCD image information, current and voltage signals, and spectral signals are collected in real time during the arc fuse additive process. The size and location information of pore defects are obtained by CT scanning of the printed part, and a pore defect feature set is analyzed and constructed. Porosity defects are classified into: no pore defects, small pore defects, and medium and large pore defects. The high-speed CCD images, current and voltage signals, and spectral signals collected at corresponding moments are then mapped to extract the pore defect feature data set.

[0010] In the third step, the directional optimization of the arc fuse additive process is performed based on the porosity defect feature dataset.

[0011] The second step is specifically as follows: first obtain high-speed CCD images, current and voltage signals, and spectral signals during the arc fuse additive manufacturing process, and at the same time build a pore defect data set of the printed part, and associate the signal characteristics with the pore defects; pre-build a standard defect recognition module, and use the sample defect characteristics to concentrate the pore defect samples to update the model parameters of the standard defect recognition module.

[0012] A porosity defect dataset for printed parts was constructed by collecting data during the arc fuse additive manufacturing process and performing an industrial CT scan on the printed parts after printing to obtain the location of the porosity defect center and the size of the porosity defect.

[0013] The signal characteristics are associated with the pore defects. Specifically, the arc area, current and voltage signal base value and peak value, and spectral characteristic line information in the high-speed CCD image are extracted and corresponded with the defect information. The characteristic value weights are set in segments to improve the online monitoring effect and accuracy. The video is recorded with a high-speed CCD, and frame-by-frame image processing is performed. The grayscale threshold is set, and the number of pixels in the arc part is calculated to reflect the size of the arc area. At the same time, the position of the arc fuse gun tip at that moment is determined and corresponds to the defect position, and also corresponds to the current and voltage value at that moment.

[0014] Construct the relationship between arc area and current and voltage. The relationship function is as follows:

[0015]

[0016] Among them, S is the arc area, k is the relationship coefficient, η is the thermal efficiency, U i is the instantaneous voltage, I i is the instantaneous current, v is the welding speed;

[0017] High-speed CCD, current, voltage, and spectral signals are collected simultaneously and correlated with the pore defects of the printed parts. Based on the different sizes of the pore defects, they are broken down into multi-sensor information corresponding to no pore defects, small pore defects, and medium and large pore defects.

[0018] The third step is specifically as follows: training a machine learning model based on the obtained pore defect feature data set, and simultaneously predicting the formation of pores online; and feeding back the predicted pore position and size to the arc fuse welding gun control interface to regulate the wire feeding speed and welding speed.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] (A) Since the present invention adopts an online monitoring method, using the spectrum, current and voltage, and high-speed CCD images monitored during the arc fuse additive manufacturing process, it is non-destructive and fast compared to traditional detection methods.

[0021] (B) Because the present invention adopts multiple sensing methods such as spectrum, current voltage, and high-speed CCD image, it reduces the disadvantages of signal instability caused by using a single signal source, and has the advantages of multiple signal sources, more reliable analysis results, and high credibility. At the same time, it uses machine learning methods, eliminating the need for manual annotation of large data sets for training.

[0022] (C) Since the present invention adopts the method of online monitoring and closed-loop control, the process parameters can be regulated by real-time monitoring results to achieve the advantage of reducing the formation of pore defects.

[0023] In summary, the present invention collects high-speed CCD image information, current and voltage signals, and spectral signals during the arc fuse additive manufacturing process, obtains a feature data set from the signals, and detects porosity defects in arc fuse additive manufacturing through machine learning, thereby achieving non-destructive and rapid detection of porosity defects in arc fuse additive manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of a method according to an embodiment of the present invention.

[0025] Figure 2 This is the high-speed CCD image processing process of an embodiment of the present invention.

[0026] Figure 3 The figures are comparison diagrams of pore changes before and after optimization of the embodiment of the present invention; (a) is the pore diagram before optimization of the embodiment; (b) is the pore diagram after optimization of the embodiment. DETAILED DESCRIPTION

[0027] The present invention is described in detail below with reference to the embodiments and accompanying drawings.

[0028] Reference Figure 1 , a method for detecting porosity defects in arc fuse additives based on multiple sensor signals, comprising the following steps:

[0029] The first step is to build an online monitoring platform, which includes an arc fuse robot arm, a high-speed CCD camera, a current and voltage acquisition module, and a spectrum acquisition module.

[0030] In the second step, high-speed CCD image information, current and voltage signals, and spectral signals are collected in real time during the arc fuse additive process. The size and location information of pore defects are obtained by CT scanning of the printed part, and a pore defect feature set is analyzed and constructed. Porosity defects are classified into: no pore defects, small pore defects, and medium and large pore defects. The high-speed CCD images, current and voltage signals, and spectral signals collected at corresponding moments are then mapped to extract the pore defect feature data set.

[0031] This embodiment first obtains high-speed CCD images, current and voltage signals, and spectral signals during the arc fuse additive manufacturing process, and simultaneously constructs a porosity defect dataset of the printed part, and associates the signal features with the porosity defects. A standard defect recognition module is pre-built, and the porosity defect samples in the sample defect feature set are used to adjust and update the model parameters of the standard defect recognition module.

[0032] The specific steps for constructing the porosity defect dataset of printed parts are as follows: data is collected during the arc fuse additive process, and after printing, the printed parts are scanned by industrial CT to obtain the location of the porosity defect center and the size of the porosity defect;

[0033] The signal features are specifically associated with pore defects as follows: the arc area, current and voltage signal base values ​​and peak values, and spectral characteristic line information are extracted from the high-speed CCD image, corresponding to the defect information, and the characteristic value weights are set in sections to improve the online monitoring effect and accuracy; high-speed CCD is used to record the video, perform frame-by-frame image processing, set the grayscale threshold, calculate the number of pixels in the arc part to reflect the size of the arc area, and simultaneously determine the position of the arc fuse gun tip at that moment, corresponding to the defect location and the current and voltage values ​​at that moment;

[0034] Construct the relationship between arc area and current and voltage. The relationship function is as follows:

[0035]

[0036] Among them, S is the arc area, k is the relationship coefficient, η is the thermal efficiency, U i is the instantaneous voltage, I i is the instantaneous current, v is the welding speed;

[0037] Simultaneously collect high-speed CCD, current, voltage, and spectral signals, correlate them with the pore defects of the printed parts, and decompose them into multi-sensor information corresponding to no pore defects, small pore defects, and medium-to-large pore defects based on the size of the pore defects;

[0038] In the third step, the directional optimization of the arc fuse additive process is performed based on the porosity defect feature dataset.

[0039] This embodiment performs machine learning model training based on the obtained pore defect feature dataset and simultaneously predicts pore formation online; based on the predicted pore position and size, feedback is fed back to the arc fuse welding gun control interface to regulate the wire feeding speed and welding speed.

[0040] Reference Figure 2 , Figure 2 The high-speed CCD image processing process of this embodiment is as follows: the video captured by the high-speed CCD is processed frame by frame, and then the initial image is gray-scaled to extract the number of pixels in the arc area to reflect the size of the arc area; at the same time, since the substrate will reflect the arc during the arc fuse additive process, it will interfere with the extraction of pixels, so it is necessary to perform regional selection on the obtained grayscale image to eliminate this influence, that is, Figure 2 The final image result is shown.

[0041] Reference Figure 3 , Figure 3 The figures are comparison diagrams of porosity changes before and after optimization of this embodiment; (a) is the porosity diagram before optimization of the embodiment, and (b) is the porosity diagram after optimization of the embodiment; it can be seen that before optimization of the embodiment, the arc fuse additive has obvious porosity defects. After optimization of the embodiment, the purpose of changing the heat input is achieved by changing the wire feeding speed and welding speed. In the additive manufacturing process, porosity defects usually increase with the increase of heat input. By reducing the heat input at the position where the porosity trend is detected, the porosity is effectively reduced.

Claims

1. A method for detecting porosity defects in arc fuse additives based on multiple sensor signals, characterized in that: The following steps are involved: The first step is to build an online monitoring platform, which includes an arc fuse robot arm, a high-speed CCD camera, a current and voltage acquisition module, and a spectrum acquisition module. In the second step, high-speed CCD image information, current and voltage signals, and spectral signals are collected in real time during the arc fuse additive process. The size and location information of pore defects are obtained by CT scanning of the printed part, and a pore defect feature set is analyzed and constructed. Porosity defects are classified into: no pore defects, small pore defects, and medium and large pore defects. The high-speed CCD images, current and voltage signals, and spectral signals collected at corresponding moments are then mapped to extract the pore defect feature data set. First, high-speed CCD images, current and voltage signals, and spectral signals are acquired during the arc fuse additive manufacturing process. A porosity defect feature set of the printed part is constructed, and the signal features are correlated with the porosity defects. Pre-build a standard defect recognition module and use pore defect samples from the pore defect feature set to update the model parameters of the standard defect recognition module; The specific steps to associate signal features with pore defects are: extracting the arc area, current and voltage signal base and peak values, and spectral characteristic line information from the high-speed CCD image, matching them with the defect information, and setting the characteristic value weights in sections to improve the online monitoring effect and accuracy; High-speed CCD is used to record video, perform frame-by-frame image processing, set grayscale thresholds, calculate the number of pixels in the arc part to reflect the size of the arc area, and simultaneously determine the position of the arc fuse gun tip at that moment, corresponding to the defect location, and also corresponding to the current and voltage values ​​at that moment; Construct the relationship between arc area and current and voltage. The relationship function is as follows: Among them, S is the arc area, k is the relationship coefficient, η is the thermal efficiency, U i is the instantaneous voltage, I i is the instantaneous current, ν is the welding speed; Simultaneously collect high-speed CCD, current, voltage, and spectral signals, correlate them with the pore defects of the printed parts, and decompose them into multi-sensor information corresponding to no pore defects, small pore defects, and medium-to-large pore defects based on the size of the pore defects; In the third step, the directional optimization of the arc fuse additive process is performed based on the porosity defect feature dataset.

2. The method according to claim 1, characterized in that The specific steps for constructing the pore defect feature set of the printed part are: collecting data during the arc fuse additive process, performing an industrial CT scan on the printed part after printing is completed, and obtaining the location of the pore defect center and the pore defect size information.

3. The method according to claim 1, characterized in that The third step is specifically as follows: training a machine learning model based on the obtained pore defect feature data set, and simultaneously predicting the formation of pores online; and feeding back the predicted pore position and size to the arc fuse welding gun control interface to regulate the wire feeding speed and welding speed.

Images