Real-time state monitoring and closed-loop control system and method for laser deposition forming process

Abstract

The invention provides a real-time state monitoring and closed-loop control system and method in the laser deposition forming process, and relates to the technical field of additive manufacturing and artificial intelligence interaction.The system comprises an acoustic emission sensor, a signal acquisition feature recognition and feedback module, a numerical control module, a laser cladding head, a powder feeder and a powder feeding pipe; wherein the acoustic emission sensor is used for acquiring an acoustic signal in the material increasing process; the signal acquisition feature recognition and feedback module is used for receiving and processing the acoustic signals received by the acoustic emission sensor, extracting acoustic signal features through a machine learning algorithm, recognizing the current machining state and feeding back the current machining state to the numerical control module; and the numerical control module is connected with the laser cladding head and the powder feeder and is used for adjusting the laser power and the powder feeding amount of the laser cladding head and the powder feeder in real time according to the processing state information. Through a linkage mechanism of sound signal real-time monitoring, feature recognition and parameter closed-loop adjustment, the stability of the material adding process can be guaranteed, and the quality of formed parts is improved.

Description

Technical Field

[0001] This invention relates to the fields of additive manufacturing and artificial intelligence interaction technology, specifically to a real-time status monitoring and closed-loop control system and method for laser deposition forming process. Background Technology

[0002] Laser deposition modeling (LDM) is a type of additive manufacturing technology that uses a laser as an energy source to simultaneously feed metal powder. Through laser melting and rapid solidification, metal powder is deposited layer by layer to create highly customizable metal structural components, widely used in aerospace and medical fields. However, due to the complexity of the LDM process and the numerous influencing parameters, even with optimal experimental parameters, defects are difficult to avoid, limiting mass production. Therefore, real-time monitoring of the product's processing status, issuing early warnings for potential defects, and timely adjustment of process parameters to achieve closed-loop control are crucial for improving product quality and manufacturing efficiency in LDM.

[0003] When materials or components deform or crack under stress, they release strain energy in the form of elastic waves, which manifest as acoustic signals on the surface of the material or component. By using acoustic emission equipment to receive the acoustic signals from the interaction between the laser and the powder phase, the current processing status of the laser deposition modeling process can be monitored. Compared with vision-based status monitoring and closed-loop control methods for the laser deposition modeling process, acoustic emission sensors are less expensive, easier to deploy, unaffected by obstructions, and more sensitive to defect formation.

[0004] Patent application CN107626923A discloses a quality control method and system for laser deposition formed products, belonging to the field of additive manufacturing technology. This method uses an infrared temperature acquisition device to collect the temperature of the starting position and surrounding surface of each scanning zone before deposition during the metal laser deposition process, quantifies and compares the data, generates instruction information for adjusting process parameters, and transmits it to a feedback control system. The feedback control system then generates different feedback instructions to adjust the laser deposition control system, thereby achieving control of the laser deposition manufacturing device. However, this patent cannot completely solve the existing technical problems, nor can it meet the needs of this invention. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a real-time status monitoring and closed-loop control system and method for laser deposition forming processes.

[0006] The real-time status monitoring and closed-loop control system for laser deposition forming process provided by the present invention includes: an acoustic emission sensor, a signal acquisition feature recognition and feedback module, a numerical control module, a laser cladding head, a powder feeder, and a powder feeding pipe; The acoustic emission sensor is used to acquire acoustic signals during the additive manufacturing process; The signal acquisition feature recognition and feedback module is used to receive and process the acoustic signal transmitted by the acoustic emission sensor, identify the current processing status, and feed it back to the CNC module. The numerical control module is connected to the laser cladding head and the powder feeder, and is used to adjust the laser power and powder feeding amount in real time according to the processing status information. The acoustic emission sensor is attached and fixed to the side wall of the substrate; The signal acquisition feature recognition and feedback module is connected to the acoustic emission sensor and the numerical control module, respectively. The laser cladding head and the powder feeder are connected by a powder feeding pipe.

[0007] Preferably, the frequency range of the acoustic emission sensor is 50~400 kHz.

[0008] The real-time status monitoring and closed-loop control method for laser deposition forming process provided by the present invention includes the following steps: Step 1: Determine multiple sets of processing parameters for laser power and powder feeding amount; Step 2: Install the acoustic emission sensor and start the processing, simultaneously acquiring acoustic signals; Step 3: Stop acquiring sound signals after processing is complete; Step 4: Perform defect analysis on the machined parts to determine their machining status; Step 5: Extract the parameters of each group of sound signals, including frequency, amplitude, energy, ring count, duration, rise time, root mean square voltage, and peak count; Step 6: Compare the acoustic signal parameters of each group and extract the characteristic parameters that reflect the processing status; Step 7: Establish a mapping model between processing parameters, processing status, and acoustic signal characteristics; Step 8: Implement closed-loop control of the processing based on the mapping model.

[0009] Preferably, step 4 includes: performing a computed tomography scan on the part, analyzing the defect type and size, and determining the processing status.

[0010] Preferably, in step 5, the time-domain parameters are calculated by analyzing the time series waveform, and the frequency-domain parameters are obtained by Fourier transform.

[0011] Preferably, in step 7, the mapping model is established through a machine learning algorithm, with inputs including laser power, powder feeding amount and acoustic signal characteristic parameters, and output being the processing status; The machine learning algorithm is the random forest algorithm.

[0012] Preferably, the machine learning algorithm constructs the mapping model in the following manner: Collect multiple sets of training samples, each with an input feature vector of... Where P is the laser power and R is the powder feed rate. to These are the n acoustic signal feature parameters extracted; The output label Y represents the corresponding processing status category; Construct a random forest model containing T decision trees, where the t-th decision tree generates a training subset from all training samples using Bootstrap during training. And at each node splitting, m features are randomly selected from all features to calculate the optimal split point, where m <n+2; When splitting a decision tree node, Gini impurity or information gain is used as the splitting criterion to determine the optimal split point. For node q, the formula for calculating its Gini impurity G(q) is: ,in The proportion of samples belonging to the i-th processing state in node q.

[0013] Preferably, when the random forest model makes predictions, the prediction results for T decision trees are... The final processing state prediction result is generated using either majority voting or average probability method. ; The decision formula for majority voting is: , where mode is the mode operator.

[0014] Preferably, the acoustic signal characteristic parameters include the frequency domain energy ratio obtained through Fourier transform, and a specific frequency band [ Energy percentage The calculation formula is:

[0015] in, This represents the frequency domain of the sound signal after Fourier transform. The sampling frequency.

[0016] Preferably, step 8 includes: adjusting the laser power and powder feeding rate in real time according to the identified processing status, wherein: if a pore defect is identified, the laser power is increased and the powder feeding rate is decreased; if a crack defect is identified, the laser power is decreased and the powder feeding rate is increased.

[0017] Compared with the prior art, the present invention has the following beneficial effects: (1) By using an acoustic emission sensor to acquire acoustic signals during the processing, and using a signal acquisition feature recognition and feedback module to extract features and identify the state through machine learning algorithms, the technical problem that traditional methods are difficult to monitor the internal process state of laser deposition forming in real time and accurately is solved, and real-time monitoring of defects such as cracks and pores that are highly sensitive and unaffected by obstruction is realized. (2) By establishing a mapping model of the three factors of processing parameters, processing status and acoustic signal characteristics, and adjusting the laser power and powder feeding amount in real time by the CNC module, the technical problem of fixed processing parameters being difficult to adapt to dynamic changes and thus causing defects is solved, and precise linkage closed-loop control of laser power and powder feeding amount is realized. (3) By extracting time-domain parameters (such as amplitude, duration, rise time, ring count, peak count, and root mean square voltage) and frequency-domain parameters (such as frequency and energy ratio) of acoustic signals based on a large amount of experimental data in the method described above as features, the technical problem that a single signal parameter is difficult to comprehensively and reliably characterize complex processing states is solved, and more accurate identification and classification of processing states (normal, pores, cracks) is achieved. (4) By using an acoustic emission sensor with a frequency range of 50~400 kHz and fixing it to the side wall of the substrate, the technical problems of low frequency band being easily affected by environmental noise and high frequency band having large signal attenuation and difficulty in effectively capturing defect features are solved, and the effective acoustic emission signal is stably and sensitively collected in a strong noise process environment. (5) By using the parameter adjustment rules in the closed-loop control method, such as “increasing the laser power or decreasing the powder feeding rate when the defect is identified as a pore defect, and decreasing the laser power or increasing the powder feeding rate when the defect is identified as a crack defect”, the technical problem of lacking an effective and directional parameter adjustment strategy to suppress defect development after abnormal conditions are detected is solved, and adaptive process control for different defect types is realized, effectively suppressing defects. Attached Figure Description

[0018] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a block diagram illustrating the control principle of the real-time status monitoring and closed-loop control method for laser deposition forming process in this invention. Figure 2 This is a diagram showing the composition of the real-time status monitoring and closed-loop control device for the laser deposition forming process in this invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0020] Example 1 This invention provides a method for real-time status monitoring and closed-loop control of a laser deposition forming process, referring to... Figure 1 As shown, the method specifically includes: Step 1: Determine the laser power and powder feed rate processing parameters for multiple laser deposition forming processes. Based on the target TC4 titanium alloy part material, the preset laser power range is 1000~1500 W, and the powder feed rate is 0.5~0.8 r / min. Sixty sets of process parameters within the parameter range are designed using orthogonal experimental design. Each set of parameters undergoes 20 repeatability tests, and each test set is uniquely numbered for subsequent data correlation.

[0021] Step 2: Based on the processing requirements, a CNC machining program is prepared. The acoustic emission sensor is installed, and the power to the acoustic signal information acquisition, recognition, and feedback module is turned on. The CNC system is then started to execute the machining program. The laser cladding head moves along a preset path, and the powder feeder feeds powder synchronously, beginning layer-by-layer deposition. The acoustic emission sensor is a piezoelectric ceramic transducer installed on the sidewall of the substrate. After acquiring the acoustic signal during the additive manufacturing process, the sensor converts it into an electrical signal, which is then connected to the signal input terminal of the signal acquisition, feature recognition, and feedback module via a wire.

[0022] Step 3: After processing is complete, stop the acquisition and recognition of acoustic signals. After processing of a single set of parameters, the CNC system automatically stops laser output and powder feeding, and the signal acquisition feature recognition and feedback module simultaneously stops signal recording and stores the acoustic signal data in the database.

[0023] Step 4: Perform computed tomography (CT) scans on the machined parts to analyze the defect types and dimensions of the parts processed by each set of machining parameters, and determine the corresponding machining state for that set of parameters. Perform CT scans on the parts processed by each set of parameters to generate a three-dimensional defect distribution map, analyze the defect types (cracks, pores) and defect sizes, and mark the machining state for each set of parameters (e.g., pore, 1.2mm).

[0024] Step 5: Using information processing algorithms, calculate the amplitude, duration, rise time, ring count, peak count, root mean square voltage, and other time-domain parameters of the sound signal by directly analyzing the time series waveform of the signal. Then, convert the time-domain signal to the frequency domain signal through Fourier transform to obtain the frequency distribution of the signal and the energy proportion of each frequency band.

[0025] Step 6: Perform correlation analysis on the acoustic signal parameters (amplitude, frequency distribution, energy ratio) and processing status (defect type, size) corresponding to multiple sets of processing parameters, calculate the correlation between each parameter and the processing status, and select the parameter with the highest correlation as the feature parameter (for example, if the "energy ratio of a certain frequency band" increases significantly in the crack state and is stable in the normal state, then it is selected as the feature parameter).

[0026] Step 7: Establish a mapping model between processing parameters, processing status, and acoustic signal features using machine learning techniques. First, construct a dataset with laser power, powder feed rate, and feature parameters as input features, and processing status as the output label, resulting in 1200 samples. Processing status is categorized into normal, abnormal (porosity), and abnormal defects (cracks). Then, use a random forest algorithm to divide the dataset into training and test sets at a 10:2 ratio. Finally, fit the mapping relationship between input and output using the training set, adjust the algorithm parameters, and ensure that the accuracy on the test set is ≥90%.

[0027] Step 8: Based on the established mapping model, the acoustic signal acquisition feature recognition and feedback module can identify the processing status of the laser deposition forming process in real time. The CNC system adjusts the laser power and powder feeding rate of the laser and powder feeder according to the changes in processing status information to achieve closed-loop control. The specific parameter adjustment rules are as follows: if a pore defect is identified, the laser power is increased and the powder feeding rate is decreased; if a crack defect is identified, the laser power is decreased and the powder feeding rate is increased. The modification ratio of additive parameters depends on the model results.

[0028] The present invention provides a real-time status monitoring and closed-loop control method for laser deposition forming process. Compared with the traditional laser deposition forming process, this method can ensure the stability of the additive process and improve the quality of the formed parts.

[0029] This invention also provides a real-time status monitoring and closed-loop control system for the laser deposition forming process. It utilizes acoustic emission equipment and its signal acquisition, feature recognition, and feedback system to identify and provide feedback on the processing status of the laser deposition forming process. The CNC system then adjusts and controls the process parameters based on the feedback signals, ensuring the laser deposition forming process is in a stable, defect-free state. This achieves precise control of laser power and powder feed rate, guarantees the stability of the additive manufacturing process, and improves part quality. (Refer to...) Figure 2 As shown, the system includes an acoustic emission sensor, a signal acquisition feature recognition and feedback module, a numerical control module, a laser cladding head, a powder feeder, a powder feeding tube, and a substrate.

[0030] The acoustic emission sensor is attached to the side wall of the substrate with high-temperature adhesive (temperature resistance ≥200℃), 5~10 cm away from the edge of the processing area, to ensure signal acquisition sensitivity, acquire the acoustic signal of the additive manufacturing process, convert it into an electrical signal, and send it to the target via wired transmission.

[0031] The signal acquisition, feature recognition, and feedback module receives and processes the electrical signals received by the acoustic emission sensor. First, it performs noise reduction processing to filter out environmental interference signals. Then, it performs signal conversion, transforming the electrical signals into digital signals for easier subsequent analysis. Finally, it extracts specific parameters based on the feature parameter types obtained from the aforementioned real-time status monitoring and closed-loop control method for the laser deposition forming process. By comparing the real-time extracted acoustic signal feature parameters with the baseline parameters in the machine learning mapping model, the current processing status is identified and communicated to the CNC module.

[0032] The CNC module is used to compare the actual processing status information with the mapping model of processing parameters, processing status and acoustic signal characteristics to obtain the process parameters to be modified, and then adjust the laser power and powder feeding amount of the laser cladding head and powder feeder.

[0033] Furthermore, to ensure the stability and sensitivity of the acoustic signal monitoring process, a 15α acoustic emission sensor was selected. Based on previous experimental results, the frequency was susceptible to environmental noise interference below 50kHz, and it was difficult to capture defect features above 400kHz. Therefore, the frequency range was 50~400 kHz.

[0034] Example 2 Example 2 is a preferred example of Example 1.

[0035] This invention provides a real-time status monitoring and closed-loop control system for a laser deposition forming process. It utilizes acoustic emission equipment and its signal acquisition, feature recognition, and feedback system to identify and provide feedback on the processing status of the laser deposition forming process. The CNC system then adjusts and controls the process parameters based on the feedback signals, ensuring the laser deposition forming process is in a stable, defect-free state. This achieves precise control of laser power and powder feed rate, ensuring the stability of the additive manufacturing process and improving part quality. (Refer to...) Figure 2 As shown, the system includes an acoustic emission sensor, a signal acquisition feature recognition and feedback module, a numerical control module, a laser cladding head, a powder feeder, a powder feeding tube, and a substrate.

[0036] Among them, the acoustic emission sensor is attached and fixed to the side wall of the substrate to acquire the acoustic signal during the additive manufacturing process.

[0037] The signal acquisition, feature recognition, and feedback module receives and processes the acoustic signals received by the acoustic emission sensor, identifies the current processing status, and informs the CNC module. The module determines the current processing status by comparing the real-time extracted acoustic signal feature parameters with the baseline parameters in the "machine learning mapping model." Specifically, the model has pre-established a correspondence between "acoustic signal features and processing status" (e.g., energy anomalies within a specific frequency range correspond to crack defects). The module inputs the real-time feature parameters into the model and outputs the identification result of the current processing status (good forming / defect present).

[0038] The CNC module is used to adjust the laser power and powder feeding amount of the laser cladding head and powder feeder in real time based on the processing status information. After receiving the processing status information from the signal acquisition, feature recognition and feedback module, the parameters are adjusted in real time according to the preset "processing status-process parameter adjustment rules" in the "mapping model". If "porosity defects" are identified (corresponding to low acoustic signal energy), the laser power is increased (to enhance the melting effect) or the powder feeding amount is reduced (to reduce unmelted powder).

[0039] The acoustic emission sensor, model 15α, operates in the frequency range of 50–400 kHz and measures the acoustic signals of machined parts, outputting a digital signal. Acoustic emission sensors require low cost, easy placement, and sensitivity to cracks. The 15α sensor is well-suited for this scenario in terms of stability and sensitivity. However, frequencies below 50 kHz may be affected by environmental noise (such as equipment vibration), and signals above 400 kHz have weak energy and are prone to attenuation, making it difficult to effectively capture defect features.

[0040] This invention also provides a method for real-time status monitoring and closed-loop control of the laser deposition forming process, referring to... Figure 1 As shown, the method specifically includes: Step 1: Determine the processing parameters, such as laser power and powder feeding amount, for multiple laser deposition forming processes.

[0041] Step 2: Develop a CNC machining program according to the processing requirements, install the acoustic emission sensor, and connect the power supply to the acoustic signal information acquisition, recognition, and feedback module before starting printing. The acoustic emission sensor is a piezoelectric ceramic transducer, installed on the sidewall of the substrate.

[0042] Printing steps: The laser deposition forming equipment executes the pre-programmed CNC machining program and starts the equipment; the laser cladding head emits a laser beam, and the powder feeder synchronously delivers metal powder to the processing area through the powder feeding pipe. The laser melts the powder and deposits it layer by layer; the acoustic emission sensor starts synchronously and begins to collect acoustic signals from the substrate and parts.

[0043] The module operates as follows during the printing process: Real-time reception: continuously receives electrical signals transmitted by the acoustic emission sensor; Real-time processing: performs noise reduction and feature parameter extraction (such as frequency, amplitude, etc.) on the signal; Real-time identification: inputs the extracted feature parameters into the mapping model to continuously judge the processing status; Real-time feedback: if an abnormal state is detected, immediately sends the information to the CNC module to trigger parameter adjustment.

[0044] Step 3: After processing is complete, stop the acquisition and recognition of sound signals.

[0045] Step 4: Perform computed tomography scans on the machined parts to analyze the defect types and degrees of the parts processed by each set of processing parameters, and determine the processing status corresponding to that set of parameters.

[0046] Step 5: Extract parameters such as frequency, amplitude, energy, ring count, duration, rise time, root mean square voltage, and peak count for each group of sound signals.

[0047] Time-domain parameters (amplitude, duration, rise time, ring count, peak count, RMS voltage): calculated by directly analyzing the time series waveform of the signal (e.g., amplitude is the peak value of the waveform, and duration is the time from the start to the end of the signal). Frequency domain parameters (frequency, energy): The time-domain signal is converted to the frequency domain by Fourier transform or wavelet transform to obtain the frequency distribution of the signal and the energy proportion of each frequency band.

[0048] The amplitude in the time domain parameter is the absolute value of the maximum deviation from zero or the reference value in the time domain waveform of the acoustic signal. Its calculation formula is: A=max(|x(t)|), where x(t) is the time domain waveform data sequence of the acoustic signal.

[0049] The root mean square voltage (RMSVoltage) in the time-domain parameters is used to characterize the average power of the acoustic signal, and its calculation formula is as follows: Where N is the total number of sampling points in the waveform of a single acoustic emission event. Let be the voltage value at the kth sampling point.

[0050] Rise time, a parameter in the time domain, is the time interval from when the acoustic signal first exceeds a threshold until it reaches its peak value. Its calculation formula is: TR = ,in, The time when the signal first exceeds the preset threshold. This refers to the time it takes for the signal to reach its peak value.

[0051] The duration in the time-domain parameters is the time interval from when the acoustic signal first exceeds the threshold to when it last falls below the threshold. Its calculation formula is as follows: ,in, The time when the signal first exceeds the preset threshold. The time when the signal last drops below a preset threshold.

[0052] The ringing count in the time domain parameters is the number of times the amplitude of the sound signal exceeds a preset threshold.

[0053] Frequency domain parameters are obtained through Fourier transform of discrete-time signal sequences. Convert to frequency domain representation The calculation formula is: = Where N is the number of sampling points and k is the frequency index. Let be the complex representation of the k-th frequency component, and its magnitude is... This represents the amplitude of that frequency component.

[0054] Step 6: Compare the parameters of each group of acoustic signals and extract the parameters that best reflect the processing status as the characteristic parameters of the acoustic signals.

[0055] The following methods were implemented through statistical analysis and comparative verification: Correlation analysis was performed on the acoustic signal parameters (frequency, amplitude, etc.) and processing states (such as defect type and severity) corresponding to multiple sets of processing parameters; the correlation between each parameter and the processing state was calculated (e.g., the fluctuation amplitude of a parameter in the defect state is much greater than that in the normal state); the parameter with the highest correlation was selected as the characteristic parameter (e.g., if the "high-frequency energy ratio" significantly increases in the crack state and is stable in the normal state, then it is selected as the characteristic parameter). The parameters were divided into two categories: normal state (stable cloud center), abnormal state (defect), and abnormal state (crack).

[0056] Step 7: Use machine learning techniques to establish a mapping model between processing parameters, processing status, and acoustic signal characteristics.

[0057] The modeling process is as follows: Data set preparation: Collect multiple sets of sample data for "processing parameters (laser power, powder feeding amount) - acoustic signal characteristic parameters - processing status (normal / abnormal)"; Data partitioning: The dataset is divided into a training set (for model learning) and a test set (for validation); Model training: Select a machine learning algorithm (such as random forest, BP neural network, etc.), take "processing parameters + acoustic signal feature parameters" as input and "processing status" as output, and train the model; Model optimization: Verify the model's accuracy using a test set, and adjust algorithm parameters (such as the number of neural network layers and the number of random forest trees) until the model can accurately predict the correspondence between "parameters-features-states".

[0058] The main supervised learning algorithms used include: Random Forest (strong in handling nonlinear relationships and suitable for multi-parameter input scenarios); Neural Network (excels at capturing the mapping relationship between complex features and states); and Support Vector Machine (suitable for small sample data and improves classification accuracy).

[0059] Specifically, the machine learning algorithm constructs the mapping model in the following manner: Collect multiple sets of training samples, each with an input feature vector of... Where P is the laser power and R is the powder feed rate. to These are the n acoustic signal feature parameters extracted; The output label Y represents the corresponding processing status category; Construct a random forest model containing T decision trees, where the t-th decision tree generates a training subset from all training samples using Bootstrap during training. And at each node splitting, m features are randomly selected from all features to calculate the optimal split point, where m <n+2; When splitting a decision tree node, Gini impurity or information gain is used as the splitting criterion to determine the optimal split point. For node q, the formula for calculating its Gini impurity G(q) is: ,in The proportion of samples belonging to the i-th processing state in node q.

[0060] When the random forest model makes predictions, the prediction results for T decision trees are... The final processing state prediction result is generated using either majority voting or average probability method. ; The decision formula for majority voting is: , where mode is the mode operator.

[0061] Preferably, the acoustic signal characteristic parameters include the frequency domain energy ratio obtained through Fourier transform, and a specific frequency band [ Energy percentage The calculation formula is:

[0062] in, This represents the frequency domain of the sound signal after Fourier transform. The sampling frequency.

[0063] Step 8: Based on the established mapping model, the acoustic signal acquisition feature recognition and feedback module can identify the processing status of the laser deposition forming process in real time. The CNC system adjusts the laser power and powder feeding amount of the laser and powder feeder according to the changes in processing status information to achieve closed-loop control.

[0064] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0065] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A real-time status monitoring and closed-loop control system for laser deposition forming process, characterized in that, include: Acoustic emission sensor, signal acquisition feature recognition and feedback module, CNC module, laser cladding head, powder feeder and powder feeding pipe; The acoustic emission sensor is used to acquire acoustic signals during the additive manufacturing process; The signal acquisition feature recognition and feedback module is used to receive and process the acoustic signal transmitted by the acoustic emission sensor, identify the current processing status, and feed it back to the CNC module. The numerical control module is connected to the laser cladding head and the powder feeder, and is used to adjust the laser power and powder feeding amount in real time according to the processing status information. The acoustic emission sensor is attached and fixed to the side wall of the substrate; The signal acquisition feature recognition and feedback module is connected to the acoustic emission sensor and the numerical control module, respectively. The laser cladding head and the powder feeder are connected by a powder feeding pipe.

2. The real-time status monitoring and closed-loop control system for laser deposition forming process according to claim 1, characterized in that, The frequency range of the acoustic emission sensor is 50~400 kHz.

3. A control method for a real-time status monitoring and closed-loop control system for a laser deposition forming process as described in claim 1 or 2, characterized in that, Includes the following steps: Step 1: Determine multiple sets of processing parameters for laser power and powder feeding amount; Step 2: Install the acoustic emission sensor and start the processing, simultaneously acquiring acoustic signals; Step 3: Stop acquiring sound signals after processing is complete; Step 4: Perform defect analysis on the machined parts to determine their machining status; Step 5: Extract the parameters of each group of sound signals, including frequency, amplitude, energy, ring count, duration, rise time, root mean square voltage, and peak count; Step 6: Compare the acoustic signal parameters of each group and extract the characteristic parameters that reflect the processing status; Step 7: Establish a mapping model between processing parameters, processing status, and acoustic signal characteristics; Step 8: Implement closed-loop control of the processing based on the mapping model.

4. The real-time status monitoring and closed-loop control method for laser deposition forming process according to claim 3, characterized in that, Step 4 includes: performing a computed tomography scan on the part, analyzing the defect type and size, and determining the processing status.

5. The real-time status monitoring and closed-loop control method for laser deposition forming process according to claim 3, characterized in that, In step 5, the time-domain parameters are calculated by analyzing the time series waveform, and the frequency-domain parameters are obtained by Fourier transform.

6. The real-time status monitoring and closed-loop control method for laser deposition forming process according to claim 3, characterized in that, In step 7, the mapping model is established through a machine learning algorithm. The inputs include laser power, powder feeding amount and acoustic signal characteristic parameters, and the output is the processing status. The machine learning algorithm is the random forest algorithm.

7. The real-time status monitoring and closed-loop control method for laser deposition forming process according to claim 3, characterized in that, The machine learning algorithm constructs the mapping model in the following manner: Collect multiple sets of training samples, each with an input feature vector of... Where P is the laser power and R is the powder feed rate. to These are the n acoustic signal feature parameters extracted; The output label Y represents the corresponding processing status category; Construct a random forest model containing T decision trees, where the t-th decision tree generates a training subset from all training samples using Bootstrap during training. And at each node splitting, m features are randomly selected from all features to calculate the optimal split point, where m <n+2; When splitting a decision tree node, Gini impurity or information gain is used as the splitting criterion to determine the optimal split point. For node q, the formula for calculating its Gini impurity G(q) is: ,in The proportion of samples belonging to the i-th processing state in node q.

8. The real-time status monitoring and closed-loop control method for laser deposition forming process according to claim 7, characterized in that, When the random forest model makes predictions, the prediction results for T decision trees are... The final processing state prediction result is generated using either majority voting or average probability method. ; The decision formula for majority voting is: , where mode is the mode operator.

9. The real-time status monitoring and closed-loop control method for laser deposition forming process according to claim 8, characterized in that, Acoustic signal characteristic parameters include the frequency domain energy ratio obtained through Fourier transform, and specific frequency bands. Energy percentage The calculation formula is: in, This represents the frequency domain of the sound signal after Fourier transform. The sampling frequency.

10. The real-time status monitoring and closed-loop control method for laser deposition forming process according to claim 3, characterized in that, Step 8 includes: adjusting the laser power and powder feeding rate in real time according to the identified processing status, wherein: if a pore defect is identified, the laser power is increased and the powder feeding rate is decreased; if a crack defect is identified, the laser power is decreased and the powder feeding rate is increased.

Images