Internal defect monitoring method for laser directed energy deposition based on semi-supervised learning

Abstract

The present invention discloses a method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning, comprising the following steps: 1: obtaining a coaxial molten pool image and preprocessing the image, and stacking the coaxial molten pool image to form a domain-level data set; labeling data: obtaining defect information of a printed deposited part through a CT image; 2: establishing a convolutional autoencoder model, and training the convolutional autoencoder model using the domain-level data set and labeled data; 3: connecting a classifier after the encoder in the convolutional autoencoder model, and using the classifier to map the low-dimensional representation output by the encoder into the category space of the sample to achieve classification of the defect features; the present invention uses the coaxial molten pool image to establish a strong understanding of the molten pool characteristics and implicit knowledge in the LDED process based on unsupervised training of the convolutional autoencoder, and finally activates the category space of the sample through a small amount of labeled data, thereby achieving semi-supervised monitoring of internal defects of laser directed energy deposition.

Description

Technical Field

[0001] The present invention belongs to the technical field of laser additive manufacturing monitoring, and in particular relates to a method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning. Background Art

[0002] Laser Directed Energy Deposition (LDED) is an advanced additive manufacturing technology that uses a high-energy laser beam to melt metal powder or wire layer by layer to form functional metal parts. Due to its high flexibility and material utilization, as well as its ability to process a variety of metal materials, LDED has been widely used in aerospace, automotive, medical and energy fields. However, despite the many advantages of LDED technology, quality control during the manufacturing process remains a major challenge facing this field.

[0003] Currently, quality control of LDED processes relies primarily on extensive labeling and experimental characterization based on specialized expertise. These methods are not only expensive and time-consuming, but also struggle to monitor all process parameters in real time. Therefore, developing an efficient, real-time, and cost-effective monitoring method is crucial.

[0004] In recent years, data-driven approaches have demonstrated tremendous potential in many engineering fields, particularly in manufacturing process monitoring and optimization. By leveraging technologies such as big data, artificial intelligence, and machine learning, data-driven approaches can extract valuable information from vast amounts of historical data, enabling real-time monitoring and quality control of complex processes. Introducing data-driven approaches into the LDED process can effectively address challenges faced by traditional methods, improving manufacturing quality and reducing production costs.

[0005] While data-driven approaches offer significant advantages in the LDED process, they also have some limitations. First, the initial deployment and training of the system requires a large amount of high-quality data, which can be difficult to obtain quickly in actual production. Second, the system's real-time performance may be limited by hardware performance when processing massive amounts of data. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning. The method uses coaxial molten pool images and establishes a strong understanding of the molten pool characteristics and implicit knowledge in the LDED process based on unsupervised training of convolutional autoencoders. Finally, the class space of samples is activated by a small amount of labeled data to achieve semi-supervised monitoring of internal defects of laser directed energy deposition.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning comprises the following steps:

[0009] Step 1: Obtain a coaxial melt pool image, preprocess the melt pool image, use a sliding time window to monitor the melt pool state within a certain time range, and stack the melt pool images within and between layers in an orderly manner according to the channel dimension to form a domain-level dataset;

[0010] Acquisition of labeled data: CT images are used to obtain the internal structure and defect information of the printed deposited parts, and the defect information is classified into three categories: negligible, medium-sized, and large;

[0011] Step 2: Build a convolutional autoencoder model and train it using the domain-level dataset and labeled data constructed in step 1.

[0012] The convolutional autoencoder model includes an L2 normalization layer, an encoder and a decoder;

[0013] The L2 normalization layer is used to adjust the data distribution, reduce the impact of abnormal samples on model training, and accelerate the gradient descent process to achieve optimal solution convergence;

[0014] The encoder is used to identify key features and defect features in the molten pool image; it includes four convolution modules, and the four convolution modules use residual connections to form an encoder;

[0015] The decoder is composed of four convolution sampling modules, which are connected in sequence;

[0016] Step 3: Decompose the encoder and decoder in the convolutional autoencoder model trained in step 2, and connect a classifier after the encoder to form a laser directed energy deposition internal defect monitoring model with semi-supervised learning. Use the classifier to map the low-dimensional representation output by the encoder into the category space of the sample to achieve defect feature classification.

[0017] Among them, when training the laser directed energy deposition internal defect monitoring model, the weight parameters of the encoder are fixed; and a small amount of labeled data is used to train the classifier.

[0018] Furthermore, preprocessing of the melt pool image refers to first converting the melt pool image in RGB format into a grayscale image with pixel values between 0 and 255, then performing ROI cropping with the melt pool point as the center, and finally further downsampling through bilinear interpolation, wherein the melt pool point refers to the brightest point in the image.

[0019] Furthermore, a threshold segmentation method is used to extract pores from CT images and calculate their areas. The pores are then associated with the melt pool image, and the defect information is divided into three categories according to the size of the pore area.

[0020] Furthermore, the convolution module includes a convolution layer, a batch normalization layer, an activation function layer and a pooling layer connected in sequence.

[0021] Furthermore, the convolution sampling module includes a convolution layer, a batch normalization layer, an activation function layer and an upsampling layer.

[0022] Furthermore, the classifier includes a classification module and an output module. The classification module includes a fully connected layer, a batch normalization layer, an activation function layer and a random inactivation layer connected in sequence. The output module is implemented using a fully connected layer.

[0023] The present invention has the following beneficial effects:

[0024] (1) Construct a domain-level dataset to characterize the quality attributes of local areas during LDED processing, taking into account spatiotemporal factors such as the dynamic changes of the melt pool over time and the mutual influence between different layers;

[0025] (2) A semi-supervised convolutional autoencoder model is established to achieve real-time monitoring of the quality of local areas during the LDED printing process with a small amount of labeled data; unsupervised training plus category space activation is used to establish a strong understanding of the melt pool characteristics of additive data, significantly reducing the need for labeled data and enhancing the accuracy and robustness of monitoring.

[0026] (3) The autoencoder structure targets the melt pool characteristics and singular samples. The morphological characteristics of the melt pool are enhanced in the deep network of the encoder, and the invariance of the model to pore scale changes is maintained by constraining the samples, making the autoencoder more suitable for the characteristics of additive manufacturing data.

[0027] (4) The semi-supervised convolutional autoencoder model was verified through experimental data; the experimental results demonstrated the effectiveness of local quality monitoring during the LDED process; by comparing with other semi-supervised methods, the reasons why LDED data is incompatible with other semi-supervised methods were deeply analyzed, thus proving the superiority of the method of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a schematic diagram of the overall framework of the model proposed in this invention.

[0029] Figure 2 Schematic diagram of printing strategy and sample construction; (a) linear reciprocating scanning strategy; (b) construction of domain-level samples.

[0030] Figure 3 Schematic diagram of the ablation experiment training process; (a) loss on the training set, (b) accuracy on the test set.

[0031] Figure 4The classification accuracy of the three region qualities under different labeled data ratios.

[0032] Figure 5 is the confusion matrix of SCAE on the test set.

[0033] Figure 6 ROC curves and AUC scores of different semi-supervised models on the test set; (a) meanteacher, (b) pimodel, (c) FixMatch, (d) SoftMatch, and (e) the SCAE model of the present invention. DETAILED DESCRIPTION

[0034] This embodiment provides a method for monitoring internal defects in laser directed energy deposition based on semi-supervised learning, comprising the following steps:

[0035] Step 1: Acquire melt pool images and construct domain-level datasets using coaxial melt pool images.

[0036] During the LDED printing process, a camera coaxially installed with the laser beam is used to obtain coaxial melt pool images in real time. The melt pool images are preprocessed, and a sliding time window is used to monitor the melt pool state within a certain time and range. The melt pool images within and between layers are stacked in an orderly manner according to the channel dimension to form a domain-level dataset.

[0037] Preprocessing the melt pool image involves converting the RGB image into a grayscale image with pixel values between 0 and 255. This is followed by cropping the ROI centered around the melt pool point, and finally downsampling using bilinear interpolation to reduce the computational burden of the deep learning process. The melt pool point is the brightest point in the image.

[0038] The domain-level dataset can reflect the dynamic melt pool information within and between layers in a local area. This embodiment uses a sampling time window with a length of 0.4s to monitor the melt pool status information within a certain time and range, and then stacks the corresponding melt pool images within the 0.4s sampling time window of each layer in an orderly manner along the channel dimension to form a domain-level dataset, reflecting the quality status of an area.

[0039] Acquisition of marked data: X-ray CT imaging technology is used to obtain the internal structure and defect information of the printed deposited parts, and to quantify the local defect information in the deposited parts.

[0040] This embodiment uses a threshold segmentation method to extract pores from CT images and calculate their areas. The pores are then associated with the melt pool image, and the defect information is divided into three categories according to the size of the pore area: negligible (marked as 0), medium-sized (marked as 1), and large (marked as 2).

[0041] Step 2: A convolutional autoencoder model is established based on the morphological characteristics and variable porosity distribution of the melt pool, and the convolutional autoencoder model is trained using the domain-level dataset constructed in step 1.

[0042] The convolutional autoencoder can reconstruct all melt pool images without the need for data labels, and learn and extract effective defect features from the original dataset; Figure 1 As shown, the convolutional autoencoder includes an L2 normalization layer, an encoder and a decoder.

[0043] Visualization of the CT scans reveals that the defect information contains some large pores, resulting in singular samples in the dataset. During training, these singular samples can cause numerical instability, particularly when smaller eigenvectors are dominated by larger ones. This implementation uses an L2 normalization layer to mitigate the impact of these singular samples on training. The samples are projected onto the surface of the unit sphere for constraint purposes, ensuring that the vectors are uniformly scaled so that each plays an equal role in calculating distance.

[0044] The key benefit of L2 normalization is its ability to enhance data invariance to scale changes. That is, regardless of the size or scale of the original data, L2 normalized data maintains a consistent distribution. This enhances numerical stability and accelerates convergence during training.

[0045] The encoder is used to identify key features and defect features in the melt pool image; this embodiment uses a residual network structure to construct the encoder, improves feature utilization efficiency, enables information to flow between different layers, and enhances the network's ability to focus on melt pool morphological features (such as edges and contours).

[0046] The encoder is composed of multiple convolution modules, each of which has the same structure. The convolution modules include a convolution layer, a batch normalization layer, an activation function layer, and a pooling layer connected in sequence. In order to more conveniently describe the structure of the encoder, this embodiment adds the first, second, third, and fourth before the convolution module to distinguish them, which has no substantive meaning. Specifically, the encoder includes a first convolution module, a second convolution module, a third convolution module, and a fourth convolution module. The first convolution module, the second convolution module, and the third convolution module are connected in sequence. The outputs of the first convolution module and the third convolution module are added and connected to the fourth convolution module.

[0047] In order to enhance the encoder's extraction of melt pool features, residual connections are used to construct the encoder, and shallow network features are merged with deep network features to enhance the morphological characteristics of the melt pool; this not only speeds up the training process, but also effectively improves the model's extraction effect on melt pool features and improves the model's classification performance.

[0048] The output of each convolution module in the encoder is xl =P(Conv l (x l-1 )), where x l-1 Represents the output of the previous convolution module, Conv l (·) is a convolution operation, which includes three consecutive transformations: convolution, batch normalization, and nonlinear activation; P(·) is a maximum pooling operation.

[0049] The decoder is used to restore image resolution through upsampling. The purpose of using the decoder in this embodiment is to complete the training of the encoder in the autoencoder, enabling it to accurately identify defect features in the melt pool image. The decoder is composed of multiple convolutional sampling modules. Each convolutional sampling module has the same structure, including a convolutional layer, a batch normalization layer, an activation function layer, and an upsampling layer. The four convolutional sampling modules are connected in sequence to output the restored image.

[0050] The output of each convolution sampling module of the decoder is x′ l =U(Conv l (x′ l-1 )), where x′ l-1 Represents the output of the previous convolution sampling module, Conv l (·) is a convolution operation, which includes three consecutive transformations: convolution, batch normalization, and nonlinear activation; U(·) is an upsampling operation.

[0051] The domain-level data in step 1 is used to train the convolutional autoencoder model. The Adam optimization algorithm and the error back propagation algorithm are used to iteratively optimize the weight parameters and bias parameters of the model to minimize the reconstruction loss function and obtain the optimized network parameters.

[0052] Step 3: Decompose the encoder and decoder in the convolutional autoencoder model trained in step 2, and connect a classifier after the encoder to form a laser directed energy deposition internal defect monitoring model with semi-supervised learning; use the classifier to map the low-dimensional representation output by the encoder into the category space of the sample to realize the classification of defect features.

[0053] The classifier includes a classification module and an output module. The classification module includes a fully connected layer, a batch normalization layer, an activation function layer and a random inactivation layer connected in sequence. The output module is implemented using a fully connected layer.

[0054] When training the laser directed energy deposition internal defect monitoring model, the encoder weight parameters are fixed; a small amount of labeled data is used to train the classifier, and the Adam optimization algorithm and error back propagation algorithm are used to iteratively optimize the model's weight parameters and bias parameters to reconstruct the loss function and obtain the optimized network parameters.

[0055] This example leverages the power of convolutional neural networks and autoencoders to extract melt pool features in an unsupervised manner. It then uses a small number of labeled samples and a classifier to introduce category information to the encoder. This approach successfully implements semi-supervised learning in the LDED monitoring process.

[0056] Experimental verification of the effectiveness of this embodiment

[0057] Experimental equipment: A fiber laser (TurDisk 4006) and a laser head (YC52) were mounted on a six-axis robotic arm (RX-160). 30CrNi2MoVA metal powder was fed into a powder feeder (RC-PGF-D). A CCD camera (CF8 / 5MX) was coaxially mounted with the laser beam, and a beam splitter and filter were placed in the optical path.

[0058] Before the experiment, the substrate was polished, and real-time melt pool images (768 × 494 pixels) were captured at 50 fps. A high-energy laser beam with a wavelength of 1030 nm was emitted by a laser, which was reflected onto the substrate by a beam splitter covered with a 1030 nm high-reflectivity transparent film. The laser and material interacted on the substrate, forming an interaction region centered on the melt pool. Light emitted from the melt pool passed through a spectrometer and reflector, and then was captured as a melt pool image by a coaxial camera.

[0059] The deposition parts were printed using different process parameters shown in Table 1. Laser power and scanning speed play a crucial role in determining the quality of the deposition parts. Therefore, various combinations of these two parameters were used to enrich the experimental data. A linear reciprocating scanning strategy was used to print the deposition parts. Each deposition part was 40 mm long and consisted of 30 layers. Figure 2 As shown, the reciprocating printing strategy helps to evenly distribute heat while reducing heat accumulation and thermal stress within the printed parts, ultimately improving printing quality and efficiency.

[0060] Table 1 Process parameters

[0061]

[0062] A domain-level dataset was constructed and trained and tested on four NVIDIA GeForce RTX 2080Ti GPUs. A total of 2,866 samples were collected and then split into training and test datasets in an 8:2 ratio. Table 2 lists some of the hyperparameters used in the model architecture. The conv2d and linear layers in the model were initialized using kaiming_normal and a normal distribution with mean zero and standard deviation 0.01, respectively. We used the Adam optimizer with a learning rate of 0.01 and the MultiStepLR strategy with a gamma value of 0.1 to adjust the learning rate.

[0063]

[0064] Ablation experiments

[0065] To intuitively demonstrate the effectiveness of residual connections and L2 normalization, a series of experiments were conducted using different model configurations: a baseline model lacking residual connections and L2 normalization, a CAE-L2 model lacking residual connections, a CAE-Res model lacking L2 normalization, and a CAE-Res-L2 model combining residual connections and L2 normalization. 20% of the labeled samples were randomly selected for training, and the performance of the model was evaluated on the test dataset. Figure 3 The training loss and test accuracy of the four models are shown. Table 3 summarizes the classification results of these models.

[0066] Table 3 Ablation experimental study of L2 normalization and residual connection

[0067]

[0068] Table 3 shows that using residual connections and L2 normalization techniques can improve prediction accuracy by 2.3% and 1.8%, respectively. Using these two techniques together improves accuracy by 3.7%, predicting the negligible state with 85% accuracy, the medium defect state with 77% accuracy, and the large defect state with 90% accuracy. These experiments demonstrate that L2 normalization can adjust the data distribution, ensure that each feature dimension has a consistent weight on the objective function, and mitigate the impact of outliers. The residual connection module enhances the model's representation and feature utilization by reusing shallow features, thereby better learning the morphological characteristics of the melt pool.

[0069] To further demonstrate the effectiveness and efficacy of L2 normalization, we studied the impact of different normalization methods on model performance. As shown in Table 4, min-max, z-score, and L2 normalization methods all produced positive results, with L2 normalization showing the greatest improvement. In contrast, L1 normalization led to a decrease in model performance. The rationale behind these results can be attributed to the characteristics of each normalization method. Min-max normalization has limited impact on outliers because it simply scales elements based on their maximum and minimum values, potentially ignoring extreme values. L1 normalization divides each element by the sum of its absolute values. When large outliers are present in a vector, this can cause some elements to approach zero, impacting the overall training process. Z-score is essentially a data normalization method that does not alter the distribution of the original data. The squared effect of outliers significantly increases the vector size, making L2 normalization particularly effective in addressing their impact. This sensitivity to outliers allows L2 normalization to improve model performance by better aligning the data distribution and mitigating their impact.

[0070] Table 4 Comparison of classification performance using different normalization methods

[0071]

[0072] Semi-supervised performance with different proportions of labeled data

[0073] In order to demonstrate the superiority of the semi-supervised learning method proposed in this example, semi-supervised and fully supervised learning models with the same network structure, depth and training hyperparameters were constructed. 458 (20%), 687 (30%), 917 (40%), 1146 (50%), 1375 (60%), 1604 (70%) and 2292 (100%) labeled samples were randomly selected for training, and the performance of the model on the test dataset was evaluated. Table 5 shows the performance indicators of the proposed semi-supervised method and the fully supervised method when trained with different proportions of labeled data. Figure 4 The prediction accuracy of three quality states under the same conditions was compared.

[0074] Table 5. Semi-supervised and fully supervised learning performance under different labeled data ratios

[0075]

[0076] As can be seen from Table 5, the semi-supervised learning method proposed in this embodiment significantly improves model performance, especially when labeled data is limited. Specifically, when only 20% of the labeled data is used, the prediction accuracy of the supervised learning method is 70.2%, while the semi-supervised learning method achieves an accuracy of 83.8%, which improves the prediction accuracy by 13.6%. In terms of the efficiency of labeled data utilization, although the labeled data is reduced by 80%, the prediction accuracy of the semi-supervised method only decreases by 6.1%. It is worth noting that the semi-supervised method achieves a performance level close to that of the fully supervised method using only 60% of the labeled data. Moreover, when all the labeled data is used, the accuracy of the semi-supervised method reaches a peak of 90.8%, exceeding the fully supervised method.

[0077] like Figure 4 As shown in the figure, the accuracy of the semi-supervised and fully supervised methods increases with the amount of labeled data. Comparing the results between the semi-supervised and fully supervised methods, the proposed semi-supervised method consistently outperforms the fully supervised learning method in all data scenarios, which proves the feasibility and effectiveness of the semi-supervised monitoring method proposed in this example in the LDED process.

[0078] like Figure 5As shown, the classification confusion matrix of the SCAE model on the test set has a prediction accuracy of 93 for the negligible state, 86 for the medium quality state, and 94 for the large defect quality state. The model has a relatively high prediction accuracy for large-scale quality states, which is conducive to achieving optimal quality control. In supervised learning, the model uses labels as a guide to force training and learning, but the labels usually include noise or incorrect labels, resulting in a decline in learning results. In contrast, semi-supervised learning methods utilize a large amount of unlabeled data, which may reduce the impact of noisy labels. The autoencoder of this embodiment extracts richer and more accurate feature representations from unlabeled raw data, effectively capturing the inherent structure and patterns in the data, thereby enhancing the robustness of the model to noise and outliers.

[0079] The semi-supervised learning method proposed in this example effectively exploits the potential features of unlabeled data and effectively utilizes the information from labeled data. This method significantly improves model performance when labeled data is limited, addresses the challenges posed by label scarcity in LDED process monitoring, and reduces the reliance on labeled data. Acquiring labeled data in LDED monitoring is time-consuming, difficult, and expensive. The semi-supervised method proposed in this article can effectively reduce the time and cost associated with LDED process monitoring and provide more accurate and reliable predictions.

[0080] Comparison with other advanced methods

[0081] The model of this embodiment (SCAE) is compared with the leading SOTA models in the field of semi-supervised learning, including MeanTacher, Pimodel, FixMatch, and SoftMatch; the same hardware and hyperparameter settings are used during the training process.

[0082] The performance of each model is shown in Table 6. The prediction accuracy, computational complexity, and number of parameters of each model are calculated. For time consumption, the average inference time of the model for 100 input data is calculated.

[0083] Table 6 Performance comparison of different semi-supervised models

[0084]

[0085] Figure 6 The ROC curve and AUC for each model are shown in Figure 2. The ROC curve evaluates the model's performance on the classification problem at different threshold levels, with curves closer to the upper left corner indicating better performance. The AUC score reflects the classifier's ability to rank samples; higher AUCs indicate better discrimination between true positives and true negatives.

[0086] Results show that the SCAE model proposed in this example achieves the highest accuracy and offers significant advantages in terms of parameter count and model inference speed (FPS). Specifically, SCAE achieves an average AUC score of 0.90, significantly higher than other models. As shown in Table 6, MeanTeacher and Pimodel achieve classification prediction accuracy averaging approximately 60%. They utilize data augmentation and consistency regularization to capture key characteristics of the target. However, LDED samples consist of a series of dynamic images; data augmentation can alter the size, brightness, and tail orientation of the melt pool, rendering the samples invalid and generating incorrect sample-label pairs. When data augmentation techniques such as random cropping and flipping are applied under full supervision, accuracy drops by 3.7%, as shown in Table 7. This demonstrates that data augmentation is not effective in the fabricated additive manufacturing dataset. FixMatch and SoftMatch are hybrid models that combine consistency regularization with pseudo-labeling to generate pseudo-labels for unlabeled samples with high confidence. However, overconfident pseudo-labels can lead to model overfitting. This is particularly problematic in datasets with noisy labels, as incorrect pseudo-labels introduce excessive noise and mislead the learning process. FixMatch achieved a prediction accuracy of 60.1%, while SoftMatch achieved an accuracy of 65.1%, which is slightly higher due to its dynamic threshold mechanism that may filter out some incorrect pseudo-labels. During training, the accuracy of labeled data quickly reached 99%, while the accuracy of unlabeled data was only 55%.

[0087] Table 7 Performance of data augmentation in fully supervised models

[0088]

[0089] The autoencoder used in this example reconstructs data directly from the raw data, establishing a deep understanding of the melt pool's characteristics and defect information without the need for additional image preprocessing or data augmentation. The overall semi-supervised learning framework also avoids the use of pseudo-labeling techniques, using only a small amount of labeled data for class space activation, effectively mitigating the impact of noisy labels.

[0090] The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any modification and replacement based on the technical solution and inventive concept provided by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for monitoring internal defects in laser directed energy deposition based on semi-supervised learning, characterized in that: The steps include: Step 1: Obtain a coaxial melt pool image, preprocess the melt pool image, use a sliding time window to monitor the melt pool state within a certain time range, and stack the melt pool images within and between layers in an orderly manner according to the channel dimension to form a domain-level dataset; Acquisition of labeled data: CT images are used to obtain the internal structure and defect information of the printed deposited parts, and the defect information is classified into three categories: negligible, medium-sized, and large; Step 2: Build a convolutional autoencoder model and train it using the domain-level dataset and labeled data constructed in step 1. The convolutional autoencoder model includes an L2 normalization layer, an encoder and a decoder; The L2 normalization layer is used to adjust the data distribution, reduce the impact of abnormal samples on model training, and accelerate the gradient descent process to achieve optimal solution convergence; The encoder is used to identify key features and defect features in the molten pool image; it includes four convolution modules, and the four convolution modules use residual connections to form an encoder; The decoder is composed of four convolution sampling modules, which are connected in sequence; Step 3: Decompose the encoder and decoder in the convolutional autoencoder model trained in step 2, and connect a classifier after the encoder to form a laser directed energy deposition internal defect monitoring model with semi-supervised learning. Use the classifier to map the low-dimensional representation output by the encoder into the category space of the sample to achieve defect feature classification. Among them, when training the laser directed energy deposition internal defect monitoring model, the weight parameters of the encoder are fixed; and a small amount of labeled data is used to train the classifier.

2. The method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning according to claim 1, characterized in that: The preprocessing of the melt pool image refers to first converting the melt pool image in RGB format into a grayscale image with pixel values between 0 and 255, then performing ROI cropping with the melt pool point as the center, and finally further downsampling through bilinear interpolation, where the melt pool point refers to the brightest point in the image.

3. The method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning according to claim 1, characterized in that: Threshold segmentation method is used to extract pores from CT images and calculate their areas; The pores are associated with the melt pool image, and the defect information is divided into three categories according to the size of the pore area.

4. The method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning according to claim 1, characterized in that: The convolution module includes a convolution layer, a batch normalization layer, an activation function layer and a pooling layer connected in sequence.

5. The method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning according to claim 1, characterized in that: The convolution sampling module includes a convolution layer, a batch normalization layer, an activation function layer and an upsampling layer.

6. The method for monitoring internal defects of laser directed energy deposition based on semi-supervised learning according to claim 1, characterized in that: The classifier includes a classification module and an output module. The classification module includes a fully connected layer, a batch normalization layer, an activation function layer and a random inactivation layer connected in sequence. The output module is implemented using a fully connected layer.

Images