A method of molten pool monitoring in an SLM melting process

Abstract

The present application relates to the technical field of molten pool monitoring, and discloses a molten pool monitoring method in an SLM melting process, which comprises the following steps: collecting molten pool image data and molten pool sound signal data in the SLM process in real time; calculating the vibration speed of the molten pool based on the molten pool sound signal data and calculating the spectral irradiance of the molten pool based on the molten pool image data; constructing a deep belief network model based on the vibration speed and the spectral irradiance of the molten pool, extracting features of the molten pool image data according to the deep belief network model to obtain first molten pool image features and second molten pool image features; searching for an optimal hyperplane to obtain a molten pool state discriminant function by combining the first molten pool image features and the second molten pool image features through a support vector machine algorithm; and identifying the molten pool image data and the molten pool sound signal data based on the molten pool state discriminant function to obtain the real-time state of the molten pool. The present application improves the real-time monitoring capability of the molten pool state, can more accurately predict the real-time state of the molten pool through real-time updating of multi-dimensional vectors, and thus can predict the quality of an SLM formed part.

Description

Technical Field

[0001] This invention relates to the field of molten pool monitoring technology, and in particular to a method for monitoring the molten pool during the SLM melting process. Background Technology

[0002] Duplex stainless steel is a high-performance stainless steel in which a body-centered cubic ferrite phase and a face-centered cubic austenite phase coexist in its solid solution structure. It combines the excellent properties of both ferritic and austenitic stainless steels and is a structural material with high strength and excellent corrosion resistance.

[0003] Currently, additive manufacturing of duplex stainless steel utilizes SLM (Selective Laser Melting), which uses a high-energy beam to melt fine metal powder layer by layer, producing precision parts with small grains and high density. However, due to the complex and rapid phase transformation process inherent in the unique duplex microstructure of duplex stainless steel during additive manufacturing, it is highly susceptible to cracking during hot forming. Therefore, it is necessary to monitor the depth and width changes of the molten pool during SLM to predict the quality of the SLM-formed parts.

[0004] Existing technologies often use CNN models to identify molten pool images and analyze them. However, CNNs are more suitable for static images and have poor recognition capabilities and low prediction accuracy for dynamic molten pool images. Summary of the Invention

[0005] In view of this, the present invention proposes a method for monitoring the molten pool during the SLM melting process. By acquiring molten pool image data and molten pool acoustic signal data in real time during the SLM process, the vibration velocity of the molten pool is calculated based on the molten pool acoustic signal data, and the spectral irradiance of the molten pool is calculated based on the molten pool image data. A deep belief network model is constructed to extract features from the molten pool image data. Based on the support vector machine algorithm and combined with the extracted image features, the optimal hyperplane is searched to obtain the molten pool state discrimination function, which improves the real-time monitoring capability of the molten pool state. Through real-time updates of multi-dimensional vectors, the state parameters of the molten pool can be predicted more accurately, and the real-time state of the molten pool can be predicted more accurately, thereby predicting the quality of the SLM formed part. This solves the problem of poor recognition capability of dynamic molten pool images in the prior art.

[0006] The real-time state of the molten pool is obtained by identifying the data based on the molten pool state discrimination function.

[0007] The technical solution of this invention is implemented as follows: This invention provides a method for monitoring the molten pool during the SLM melting process, comprising the following steps:

[0008] S1, real-time acquisition of molten pool image data and molten pool acoustic signal data during the SLM process;

[0009] S2, calculate the vibration velocity of the molten pool based on the acoustic signal data of the molten pool, and calculate the spectral irradiance of the molten pool based on the image data of the molten pool;

[0010] S3. A deep belief network model is constructed based on the vibration velocity and spectral irradiance of the molten pool. Feature extraction is performed on the molten pool image data according to the deep belief network model to obtain the first molten pool image features and the second molten pool image features.

[0011] S4. By combining the features of the first molten pool image and the features of the second molten pool image with the support vector machine algorithm, the optimal hyperplane is searched to obtain the molten pool state discrimination function.

[0012] S5. Based on the melt pool state discrimination function, the melt pool image data and melt pool acoustic signal data are identified to obtain the real-time state of the melt pool, and the quality of the SLM formed part is predicted based on the real-time state of the melt pool.

[0013] Based on the above technical solutions, preferably, step S1 includes:

[0014] By arranging optical and acoustic sensors around the molten pool, image data of the molten pool during the SLM process is acquired based on the optical sensors, and acoustic signal data of the molten pool during the SLM process is acquired based on the acoustic sensors.

[0015] Based on the above technical solutions, preferably, step S1 further includes:

[0016] The molten pool image data includes video images, molten pool optical signal wavelength, and molten pool coarse measurement temperature. The molten pool acoustic signal data includes acoustic impedance and molten pool acoustic wave velocity. The optical signal wavelength and the molten pool coarse measurement temperature are used to calculate the spectral irradiance of the molten pool, and the acoustic impedance and molten pool acoustic wave velocity are used to calculate the vibration velocity of the molten pool.

[0017] Based on the above technical solutions, preferably, step S2 includes:

[0018] The vibration velocity of the molten pool is calculated based on the acoustic signal data of the molten pool, and the spectral irradiance of the molten pool is calculated based on the image data of the molten pool. The calculation formulas for the vibration velocity and spectral irradiance of the molten pool are as follows:

[0019]

[0020] Where v is the vibration velocity of the molten pool, α is the laser absorption coefficient, β is the maximum energy distribution coefficient, γ is the gas adiabatic index, and v 声 Let λ be the acoustic velocity of the molten pool, Z be the acoustic impedance, f be the spectral irradiance of the molten pool, h and c be Planck's constants, k be Boltzmann's constant, λ be the wavelength of the optical signal in the molten pool, T be the rough temperature of the molten pool, and σ be the optical sensor error bias.

[0021] Based on the above technical solutions, preferably, step S3 includes:

[0022] S31, an initial deep belief network model is constructed based on the vibration velocity and spectral irradiance of the molten pool, historical molten pool image data is obtained, and the historical molten pool image data is divided into training set and validation set;

[0023] S32, iteratively train the initial deep belief network model using the training set, configure the loss function and hyperparameters, until the iteration stopping condition is met, and then the trained deep belief network model is obtained.

[0024] S33. Use the validation set to test the trained deep belief network model and evaluate the model. If the model evaluation does not meet the requirements, return to step S32 to adjust the hyperparameters and retrain the model iteratively.

[0025] Based on the above technical solutions, preferably, step S31 includes:

[0026] Based on the vibration velocity and spectral irradiance of the molten pool, the input layer structure of the initial deep belief network model is determined, the network weights and biases of the initial deep belief network model are initialized, historical molten pool image data is collected and preprocessed, and the preprocessed historical molten pool image data is divided into training set and validation set with a ratio of 7:3. The hidden layer structure of the initial deep belief network model is determined, including the number of hidden layers and the number of neurons in each hidden layer.

[0027] Based on the above technical solutions, preferably, step S32 includes:

[0028] The initial hyperparameters are set, including the number of pre-training iterations and the number of layers in the restricted Boltzmann machine. The initial deep belief network model adopts the cross-entropy loss function and the Adam optimization algorithm. For each training batch, the network output is calculated through forward propagation, and it is checked whether the iteration stopping condition is met. The iteration stopping condition includes reaching the maximum number of iterations and the convergence of the loss function.

[0029] Based on the above technical solutions, preferably, step S33 includes:

[0030] The mAP value of the trained deep belief network model is calculated on the validation set. The current mAP value is compared with the preset mAP value. When the current mAP value is not less than the preset mAP value, the model evaluation is completed and the deep belief network model is obtained.

[0031] Based on the above technical solutions, preferably, step S4 includes:

[0032] The formula for calculating the molten pool state discrimination function is:

[0033]

[0034] Where f(x) is the molten pool state discrimination function, sign(·) is the signal function, and x i Let η be the i-th sample of the optimal hyperplane, D be the number of samples in the optimal hyperplane, and η be the number of samples in the optimal hyperplane. i Let y be the Lagrange multiplier of the i-th sample of the optimal hyperplane. i For sample x i The observed value, G(x) i (x) is the kernel function, and b1 and b2 are the biases of the first molten pool image features and the second molten pool image features, respectively.

[0035] Based on the above technical solutions, preferably, step S5 includes:

[0036] Based on the molten pool state discrimination function, the molten pool image data and molten pool acoustic signal data are identified to obtain molten pool state identification parameters. The molten pool state identification parameters include molten pool predicted depth, molten pool predicted width, molten pool predicted temperature, molten pool predicted vibration frequency, and molten pool radiation intensity. A multi-dimensional vector is constructed based on the molten pool predicted depth, molten pool predicted width, molten pool predicted temperature, molten pool predicted vibration frequency, and molten pool radiation intensity. By updating the molten pool image data and molten pool acoustic signal data in real time, the multi-dimensional vector is updated in real time to obtain the real-time state of the molten pool. Based on the real-time state of the molten pool, the quality of the SLM formed part is predicted.

[0037] The molten pool monitoring method of the present invention during the SLM melting process has the following advantages over the prior art:

[0038] (1) By collecting molten pool image data and molten pool acoustic signal data in real time during the SLM process, the vibration velocity of the molten pool is calculated based on the molten pool acoustic signal data, and the spectral irradiance of the molten pool is calculated based on the molten pool image data. A deep belief network model is constructed, and feature extraction is performed on the molten pool image data. Based on the support vector machine algorithm and the extracted image features, the optimal hyperplane is searched to obtain the molten pool state discrimination function, which improves the real-time monitoring capability of the molten pool state. Through the real-time update of multi-dimensional vectors, the real-time state of the molten pool can be predicted more accurately, thereby predicting the quality of the SLM formed parts.

[0039] (2) By combining the deep belief network model and the support vector machine algorithm, a comprehensive and accurate identification of the molten pool state was achieved, which improved the accuracy and reliability of molten pool state monitoring.

[0040] (3) By updating the molten pool image data and acoustic signal data in real time, a multidimensional vector is constructed and updated, realizing dynamic monitoring and real-time adjustment of the molten pool state. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart of a method for monitoring the molten pool during the SLM melting process according to the present invention. Detailed Implementation

[0043] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0044] Please see Figure 1 This embodiment provides a method for monitoring the molten pool during the SLM melting process, including the following steps:

[0045] S1, real-time acquisition of molten pool image data and molten pool acoustic signal data during the SLM process;

[0046] S2, calculate the vibration velocity of the molten pool based on the acoustic signal data of the molten pool, and calculate the spectral irradiance of the molten pool based on the image data of the molten pool;

[0047] S3. A deep belief network model is constructed based on the vibration velocity and spectral irradiance of the molten pool. Feature extraction is performed on the molten pool image data according to the deep belief network model to obtain the first molten pool image features and the second molten pool image features.

[0048] S4. By combining the features of the first molten pool image and the features of the second molten pool image with the support vector machine algorithm, the optimal hyperplane is searched to obtain the molten pool state discrimination function.

[0049] S5. Based on the melt pool state discrimination function, the melt pool image data and melt pool acoustic signal data are identified to obtain the real-time state of the melt pool, and the quality of the SLM formed part is predicted based on the real-time state of the melt pool.

[0050] Specifically, this embodiment of a molten pool monitoring method during the SLM melting process acquires molten pool image data and molten pool acoustic signal data in real time during the SLM process. Based on the molten pool acoustic signal data, the vibration velocity of the molten pool is calculated, and based on the molten pool image data, the spectral irradiance of the molten pool is calculated. A deep belief network model is constructed, and features are extracted from the molten pool image data. Based on the support vector machine algorithm and combined with the extracted image features, the optimal hyperplane is searched to obtain the molten pool state discrimination function, which improves the real-time monitoring capability of the molten pool state. Through the real-time update of multi-dimensional vectors, the state parameters of the molten pool can be predicted more accurately.

[0051] Step S1 includes:

[0052] By arranging optical and acoustic sensors around the molten pool, image data of the molten pool during the SLM process is acquired based on the optical sensors, and acoustic signal data of the molten pool during the SLM process is acquired based on the acoustic sensors.

[0053] Specifically, this embodiment enables the acquisition of image data and acoustic signal data of the molten pool by simultaneously using optical and acoustic sensors, thus achieving multi-dimensional and multi-angle information acquisition.

[0054] Optical sensors can capture the visual characteristics of the molten pool, while acoustic sensors can capture its acoustic characteristics. The combination of the two types of sensors makes data acquisition more comprehensive and accurate.

[0055] By placing sensors around the molten pool, molten pool data can be collected in real time during the SLM process, providing a basis for real-time analysis and monitoring.

[0056] Different types of sensors can capture different types of information. This combination enables this embodiment to adapt to more process parameters and material types. The combination of molten pool image data and molten pool acoustic signal data provides a richer source of information for deep learning and feature extraction.

[0057] Step S1 also includes:

[0058] The molten pool image data includes video images, molten pool optical signal wavelength, and molten pool coarse measurement temperature. The molten pool acoustic signal data includes acoustic impedance and molten pool acoustic wave velocity. The optical signal wavelength and the molten pool coarse measurement temperature are used to calculate the spectral irradiance of the molten pool, and the acoustic impedance and molten pool acoustic wave velocity are used to calculate the vibration velocity of the molten pool.

[0059] Specifically, this embodiment obtains the raw data required to calculate key physical quantities by acquiring video images of the molten pool, optical signal wavelength, coarse temperature measurement, acoustic impedance, and sound wave velocity. The acquired optical signal wavelength and coarse temperature measurement can be directly used to calculate the spectral irradiance of the molten pool, and the acoustic impedance and sound wave velocity can be directly used to calculate the vibration velocity of the molten pool.

[0060] By simultaneously acquiring visual, spectral, and acoustic data, the state of the molten pool can be characterized from multiple dimensions, enhancing the ability to describe the characteristics of the molten pool. By collecting optical signal wavelengths and sound wave velocities, the sensitivity of the monitoring system to minute changes in the molten pool is improved.

[0061] Step S2 includes:

[0062] The vibration velocity of the molten pool is calculated based on the acoustic signal data of the molten pool, and the spectral irradiance of the molten pool is calculated based on the image data of the molten pool. The calculation formulas for the vibration velocity and spectral irradiance of the molten pool are as follows:

[0063]

[0064] Where v is the vibration velocity of the molten pool, α is the laser absorption coefficient, β is the maximum energy distribution coefficient, γ is the gas adiabatic index, and v 声 Let λ be the acoustic velocity of the molten pool, Z be the acoustic impedance, f be the spectral irradiance of the molten pool, h and c be Planck's constants, k be Boltzmann's constant, λ be the wavelength of the optical signal in the molten pool, T be the rough temperature of the molten pool, and σ be the optical sensor error bias.

[0065] Specifically, this embodiment calculates the vibration velocity of the molten pool by using acoustic signal data of the molten pool, combined with acoustic wave velocity and acoustic impedance, thereby accurately reflecting the dynamic characteristics of the molten pool; and calculates the spectral irradiance of the molten pool by using image data of the molten pool, combined with optical signal wavelength and coarse temperature measurement, thereby accurately reflecting the thermal radiation characteristics of the molten pool.

[0066] This embodiment calculates the vibration velocity of the molten pool by considering other influencing factors such as the laser absorption coefficient, the maximum energy distribution coefficient, and the gas adiabatic index, thereby improving the accuracy of molten pool condition monitoring. By considering the optical sensor error bias, the impact of measurement errors on the calculation results can be reduced, thus improving the accuracy of the data.

[0067] Step S3 includes:

[0068] S31, an initial deep belief network model is constructed based on the vibration velocity and spectral irradiance of the molten pool, historical molten pool image data is obtained, and the historical molten pool image data is divided into training set and validation set;

[0069] S32, iteratively train the initial deep belief network model using the training set, configure the loss function and hyperparameters, until the iteration stopping condition is met, and then the trained deep belief network model is obtained.

[0070] S33. Use the validation set to test the trained deep belief network model and evaluate the model. If the model evaluation does not meet the requirements, return to step S32 to adjust the hyperparameters and retrain the model iteratively.

[0071] Specifically, this embodiment uses historical melt pool image data for training and validation. The deep belief network model can more accurately extract and identify the features of the melt pool, improving the prediction accuracy and anti-interference ability of the model in the actual SLM process.

[0072] Through iterative training and hyperparameter tuning, the model can adaptively optimize its structure and parameters, ensuring good performance under different process parameters and material conditions, thus improving the model's adaptability.

[0073] By dividing the data into training and validation sets and evaluating the model, we ensure that the model not only performs well on the training data but also maintains high performance under unseen SLM process conditions, thereby enhancing the model's ability to be applied in actual production.

[0074] Based on validation set testing and model evaluation, an effective feedback mechanism was established, enabling the model to be adjusted and improved in a timely manner according to the evaluation results, ensuring that the model achieves the expected performance standards in practical applications.

[0075] By utilizing the multi-layered structure of deep belief networks, complex nonlinear features can be extracted from vibration velocity and spectral irradiance, providing richer and more accurate information for subsequent molten pool state determination and improving the accuracy of SLM process monitoring.

[0076] Step S31 includes:

[0077] Based on the vibration velocity and spectral irradiance of the molten pool, the input layer structure of the initial deep belief network model is determined, the network weights and biases of the initial deep belief network model are initialized, historical molten pool image data is collected and preprocessed, and the preprocessed historical molten pool image data is divided into training set and validation set with a ratio of 7:3. The hidden layer structure of the initial deep belief network model is determined, including the number of hidden layers and the number of neurons in each hidden layer.

[0078] Specifically, this embodiment optimizes the network structure based on the vibration velocity and spectral irradiance of the molten pool, enabling the model to more accurately capture the characteristics of the molten pool during the SLM process, thus significantly improving the accuracy of molten pool condition monitoring.

[0079] Reasonable initialization of network weights and biases significantly shortens model training time, enabling the model to be applied to actual production more quickly and improving production efficiency. Preprocessing of historical melt pool image data effectively removes interference factors and provides more reliable training data, thereby improving the stability of the model in practical applications.

[0080] By adopting a 7:3 ratio of training to validation sets, the model can maintain good performance under different SLM process parameters and material conditions, enhancing its adaptability in actual production. By modifying the hidden layer structure, the model can better extract the complex features of the SLM melt pool, improve the ability to identify abnormal states, and help to detect and prevent potential quality problems in a timely manner.

[0081] Step S32 includes:

[0082] The initial hyperparameters are set, including the number of pre-training iterations and the number of layers in the restricted Boltzmann machine. The initial deep belief network model adopts the cross-entropy loss function and the Adam optimization algorithm. For each training batch, the network output is calculated through forward propagation, and it is checked whether the iteration stopping condition is met. The iteration stopping condition includes reaching the maximum number of iterations and the convergence of the loss function.

[0083] Specifically, this embodiment provides a good starting point for the model by setting initial hyperparameters, such as the number of pre-training iterations and the number of layers of the restricted Boltzmann machine, which helps to quickly achieve a high performance level in the SLM melt pool monitoring task.

[0084] By employing the Adam optimization algorithm, the learning rate can be adaptively adjusted, which accelerates the convergence speed of the model, reduces the time required for model training, and enables the model to be applied to actual production more quickly.

[0085] Using the cross-entropy loss function can avoid the gradient vanishing problem, making the model more stable during training and improving the reliability of the final model in actual SLM process monitoring.

[0086] By setting iteration stopping conditions, including the maximum number of iterations and the convergence criterion of the loss function, precise control of model training is ensured, overfitting or underfitting is avoided, and the generalization ability of the model in practical applications is improved.

[0087] Through batch training and dynamic adjustment, the model can better adapt to various complex situations in the SLM process, and improve its ability to identify the state of the molten pool under different process parameters and material conditions.

[0088] Step S33 includes:

[0089] The mAP value of the trained deep belief network model is calculated on the validation set. The current mAP value is compared with the preset mAP value. When the current mAP value is not less than the preset mAP value, the model evaluation is completed and the deep belief network model is obtained.

[0090] Specifically, this embodiment evaluates the model using the mAP value, which is calculated as follows:

[0091]

[0092] Where d is the number of data categories in the deep belief network model, R(t) is the recall rate at the current timestamp, R(t+1) is the recall rate at the next timestamp, and precision(t,t+1) is the precision at the current timestamp and the next timestamp.

[0093] The following parameters are specified:

[0094] TP: The number of samples correctly classified as positive; samples that are actually positive are also classified as positive by the model.

[0095] FP: The number of samples that are misclassified as positive; samples that are actually negative but are classified as positive by the model.

[0096] TN: The number of samples correctly classified as negative; samples that are actually negative are also classified as negative by the model.

[0097] FN: The number of samples that are misclassified as negative; samples that are actually positive but are classified as negative by the model.

[0098] Precision = TP / (TP+FP), Recall = TP / (TP+FN);

[0099] Using each type of Precision and Recall as the horizontal and vertical axes respectively, calculating the AP value is essentially equivalent to calculating the area enclosed by Precision and Recall. Since the resulting curve region fluctuates significantly, interpolation is used to smooth the curve. If the current recall value is R(t), the interpolation value is the maximum Precision value between the current position and the next position where the recall is R(t+1), where t represents the current position. The mAP value is the average of the AP values ​​for all types, where d represents the number of data categories in the deep belief network model.

[0100] This embodiment comprehensively evaluates the model's detection and classification capabilities across different categories by calculating the mAP value, ensuring that the model maintains a high level of performance under various melt pool conditions. Using the mAP value as an evaluation criterion can effectively identify the performance differences of the model across different categories, ensuring that the final selected model has high accuracy and reliability in practical applications.

[0101] By comparing the current mAP value with the preset mAP value, it is possible to quickly determine whether the model has reached the expected performance standard, help select the optimal model configuration, and improve the application effect of the model in SLM melt pool monitoring. If the current mAP value does not reach the preset standard, it can provide a clear direction and basis for further optimization of the model, help adjust the model parameters and structure to improve performance. By interpolating and smoothing the Precision and Recall curves, the impact of curve jitter on the evaluation results is reduced, and the accuracy of mAP value calculation is improved.

[0102] Step S4 includes:

[0103] The formula for calculating the molten pool state discrimination function is:

[0104]

[0105] Where f(x) is the molten pool state discrimination function, sign(·) is the signal function, and x i Let η be the i-th sample of the optimal hyperplane, D be the number of samples in the optimal hyperplane, and η be the number of samples in the optimal hyperplane. i Let y be the Lagrange multiplier of the i-th sample of the optimal hyperplane. i For sample x i The observed value, G(x) i (x) is the kernel function, and b1 and b2 are the biases of the first molten pool image features and the second molten pool image features, respectively.

[0106] Specifically, this embodiment uses a support vector machine algorithm to construct a melt pool state discrimination function, which achieves high-precision classification of the melt pool state during the SLM process, thereby effectively distinguishing between normal and abnormal melt pool states and improving the quality control capability in the SLM manufacturing process.

[0107] By introducing the kernel function G(x) i The addition of ,x) enables the model to perform nonlinear classification in a high-dimensional feature space, enhancing its ability to handle complex molten pool state characteristics and adapting it to various complex process parameters and material conditions in the SLM process.

[0108] By using the concept of the optimal hyperplane, the model can find the optimal decision boundary in the feature space, which improves the accuracy of melt pool state classification, enhances the model's robustness to boundary conditions, and reduces the possibility of misjudgment.

[0109] By introducing two bias terms, b1 and b2, the model can flexibly process feature information from different sources (such as features from the first and second melt pool images). This improves the model's ability to utilize multi-source data, making state discrimination more comprehensive and accurate.

[0110] Based on this discriminant function, the system can make real-time judgments on the state of the molten pool during the SLM process. This real-time monitoring capability is crucial for timely detection and correction of abnormal states in the manufacturing process, effectively improving production efficiency and product quality.

[0111] By using Lagrange multipliers η i and sample observations y i The model is better able to handle the problem of imbalanced samples and effectively reduces the false alarm rate.

[0112] Step S5 includes:

[0113] Based on the molten pool state discrimination function, the molten pool image data and molten pool acoustic signal data are identified to obtain molten pool state identification parameters. The molten pool state identification parameters include molten pool predicted depth, molten pool predicted width, molten pool predicted temperature, molten pool predicted vibration frequency, and molten pool radiation intensity. A multi-dimensional vector is constructed based on the molten pool predicted depth, molten pool predicted width, molten pool predicted temperature, molten pool predicted vibration frequency, and molten pool radiation intensity. By updating the molten pool image data and molten pool acoustic signal data in real time, the multi-dimensional vector is updated in real time to obtain the real-time state of the molten pool. Based on the real-time state of the molten pool, the quality of the SLM formed part is predicted.

[0114] Specifically, this embodiment uses a molten pool state discrimination function, combined with molten pool image data and acoustic signal data, to comprehensively identify various state parameters of the molten pool, including depth, width, temperature, vibration frequency, and radiation intensity, providing all-round monitoring of the molten pool state.

[0115] By updating the constructed multidimensional vector in real time, the changes in the molten pool state can be dynamically captured, improving the prediction accuracy of the molten pool state and thus more accurately reflecting the actual situation in the SLM process. By updating the molten pool image and acoustic signal data in real time, the system can monitor the molten pool state in real time, thereby promptly detecting and correcting abnormal states in the manufacturing process. Based on the real-time molten pool state, the quality of SLM formed parts can be predicted more accurately, which helps to identify potential quality problems in advance and reduce scrap rate and rework costs.

[0116] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the molten pool during the SLM melting process, characterized in that, Includes the following steps: S1, real-time acquisition of molten pool image data and molten pool acoustic signal data during the SLM process; S2, calculate the vibration velocity of the molten pool based on the acoustic signal data of the molten pool, and calculate the spectral irradiance of the molten pool based on the image data of the molten pool; S3. A deep belief network model is constructed based on the vibration velocity and spectral irradiance of the molten pool. Feature extraction is performed on the molten pool image data according to the deep belief network model to obtain the first molten pool image features and the second molten pool image features. S4. By combining the features of the first molten pool image and the features of the second molten pool image with the support vector machine algorithm, the optimal hyperplane is searched to obtain the molten pool state discrimination function. S5. Based on the melt pool state discrimination function, the melt pool image data and melt pool acoustic signal data are identified to obtain the real-time state of the melt pool, and the quality of the SLM formed part is predicted based on the real-time state of the melt pool.

2. The method for monitoring the molten pool during the SLM melting process as described in claim 1, characterized in that, Step S1 includes: By arranging optical and acoustic sensors around the molten pool, image data of the molten pool during the SLM process is acquired based on the optical sensors, and acoustic signal data of the molten pool during the SLM process is acquired based on the acoustic sensors.

3. The method for monitoring the molten pool during the SLM melting process as described in claim 2, characterized in that, Step S1 also includes: The molten pool image data includes video images, molten pool optical signal wavelength, and molten pool coarse measurement temperature. The molten pool acoustic signal data includes acoustic impedance and molten pool acoustic wave velocity. The optical signal wavelength and the molten pool coarse measurement temperature are used to calculate the spectral irradiance of the molten pool, and the acoustic impedance and molten pool acoustic wave velocity are used to calculate the vibration velocity of the molten pool.

4. The method for monitoring the molten pool during the SLM melting process as described in claim 3, characterized in that, Step S2 includes: The vibration velocity of the molten pool is calculated based on the acoustic signal data of the molten pool, and the spectral irradiance of the molten pool is calculated based on the image data of the molten pool. The calculation formulas for the vibration velocity and spectral irradiance of the molten pool are as follows: Where v is the vibration velocity of the molten pool, α is the laser absorption coefficient, β is the maximum energy distribution coefficient, γ is the gas adiabatic index, and v 声 Let λ be the acoustic velocity of the molten pool, Z be the acoustic impedance, f be the spectral irradiance of the molten pool, h and c be Planck's constants, k be Boltzmann's constant, λ be the wavelength of the optical signal in the molten pool, T be the rough temperature of the molten pool, and σ be the optical sensor error bias.

5. The method for monitoring the molten pool during the SLM melting process as described in claim 4, characterized in that, Step S3 includes: S31, an initial deep belief network model is constructed based on the vibration velocity and spectral irradiance of the molten pool, historical molten pool image data is obtained, and the historical molten pool image data is divided into training set and validation set; S32, iteratively train the initial deep belief network model using the training set, configure the loss function and hyperparameters, until the iteration stopping condition is met, and then the trained deep belief network model is obtained. S33. Use the validation set to test the trained deep belief network model and evaluate the model. If the model evaluation does not meet the requirements, return to step S32 to adjust the hyperparameters and retrain the model iteratively.

6. The method for monitoring the molten pool during the SLM melting process as described in claim 5, characterized in that, Step S31 includes: Based on the vibration velocity and spectral irradiance of the molten pool, the input layer structure of the initial deep belief network model is determined, the network weights and biases of the initial deep belief network model are initialized, historical molten pool image data is collected and preprocessed, and the preprocessed historical molten pool image data is divided into training set and validation set with a ratio of 7:

3. The hidden layer structure of the initial deep belief network model is determined, including the number of hidden layers and the number of neurons in each hidden layer.

7. The method for monitoring the molten pool during the SLM melting process as described in claim 6, characterized in that, Step S32 includes: The initial hyperparameters are set, including the number of pre-training iterations and the number of layers in the restricted Boltzmann machine. The initial deep belief network model adopts the cross-entropy loss function and the Adam optimization algorithm. For each training batch, the network output is calculated through forward propagation, and it is checked whether the iteration stopping condition is met. The iteration stopping condition includes reaching the maximum number of iterations and the convergence of the loss function.

8. The method for monitoring the molten pool during the SLM melting process as described in claim 7, characterized in that, Step S33 includes: The mAP value of the trained deep belief network model is calculated on the validation set. The current mAP value is compared with the preset mAP value. When the current mAP value is not less than the preset mAP value, the model evaluation is completed and the deep belief network model is obtained.

9. The method for monitoring the molten pool during the SLM melting process as described in claim 8, characterized in that, Step S4 includes: The formula for calculating the molten pool state discrimination function is: Where f(x) is the molten pool state discrimination function, sign(·) is the signal function, and x i Let η be the i-th sample of the optimal hyperplane, D be the number of samples in the optimal hyperplane, and η be the number of samples in the optimal hyperplane. i Let y be the Lagrange multiplier of the i-th sample of the optimal hyperplane. i For sample x i The observed value, G(x) i (x) is the kernel function, and b1 and b2 are the biases of the first molten pool image features and the second molten pool image features, respectively.

10. The method for monitoring the molten pool during the SLM melting process as described in claim 9, characterized in that, Step S5 includes: Based on the molten pool state discrimination function, the molten pool image data and molten pool acoustic signal data are identified to obtain molten pool state identification parameters. The molten pool state identification parameters include molten pool predicted depth, molten pool predicted width, molten pool predicted temperature, molten pool predicted vibration frequency, and molten pool radiation intensity. A multi-dimensional vector is constructed based on the molten pool predicted depth, molten pool predicted width, molten pool predicted temperature, molten pool predicted vibration frequency, and molten pool radiation intensity. By updating the molten pool image data and molten pool acoustic signal data in real time, the multi-dimensional vector is updated in real time to obtain the real-time state of the molten pool. Based on the real-time state of the molten pool, the quality of the SLM formed part is predicted.

Images