An AI-driven method for predicting faults in stereolithography 3D printing

Abstract

This invention provides an AI-driven method for predicting faults in photopolymer 3D printing, relating to the field of 3D printing technology. This method relies on sensors installed on the photopolymer 3D printing equipment to collect printing parameters in real time and perform data preprocessing; extracting multidimensional features of the photopolymer 3D printing, performing temporal and spatial feature analysis to obtain a high-dimensional feature vector; labeling collected historical printing data as normal data samples and fault data samples, and training and optimizing the constructed fault prediction model based on the high-dimensional feature vector; detecting the real-time collected printing parameters based on the trained fault prediction model, comparing the obtained prediction results with a preset warning threshold, and issuing a warning based on the fault classification result if the threshold is exceeded. This invention improves the reliability of the printing process and product quality by real-time monitoring and analysis of key parameters in the printing process, predicting potential faults, and automatically adjusting when anomalies occur.

Description

Technical Field

[0001] This invention relates to the field of 3D printing technology, and more specifically to an AI-driven method for predicting faults in photopolymerization 3D printing. Background Technology

[0002] Stereolithography (SLA) or digital light processing (DLP) is an additive manufacturing technology based on the curing of photosensitive resin. Its basic principle is to use a light source (such as a laser or projector) to irradiate the photosensitive resin, causing the resin to solidify layer by layer through a photocuring reaction, thereby gradually building a three-dimensional object. This technology is widely used in industrial manufacturing, medical, jewelry, and construction fields due to its high precision and smooth surface finish.

[0003] In photopolymer 3D printing, problems such as poor layer adhesion, dimensional deviations, or surface defects frequently occur, leading to unstable print quality. Currently, print quality is mainly improved by adjusting printing parameters (such as light source intensity, printing speed, and layer thickness). However, this relies heavily on user experience and repeated experimentation, and is difficult to handle dynamically changing printing conditions. Current photopolymer 3D printing methods focus on post-printing repair or parameter adjustment, with limited research on real-time monitoring and fault prediction. Due to the multivariate and complex nature of the photopolymer 3D printing process, existing technologies struggle to achieve comprehensive and accurate fault prediction and real-time adjustment.

[0004] Therefore, there is an urgent need to develop a fault prediction method for photopolymer 3D printing that can monitor and predict faults in real time, so as to improve the stability and reliability of photopolymer 3D printing. Summary of the Invention

[0005] In view of this, this invention proposes an AI-driven fault prediction method for photopolymer 3D printing. By real-time monitoring and analysis of key parameters during the printing process, it predicts potential faults and automatically adjusts the system when anomalies occur. Furthermore, by constructing and applying a fault prediction model, it performs real-time analysis of key parameters such as temperature, light intensity, and material flow during the photopolymer 3D printing process. The system can predict potential faults before printing begins and monitor key indicators during the printing process. For detected anomalies, the system automatically adjusts printing parameters or issues an alarm, thereby improving the reliability of the printing process and product quality.

[0006] This invention is achieved using the following technical solution:

[0007] In a first aspect, the present invention provides an AI-driven method for predicting failures in photopolymerization 3D printing, the method comprising the following steps:

[0008] The printing parameters are collected in real time by sensors installed on the photopolymerization 3D printing equipment, and the collected printing parameters are preprocessed by the data processing module to eliminate noise and outliers.

[0009] Multidimensional features of photopolymer 3D printing are extracted from the preprocessed data, and temporal and spatial features are analyzed to obtain high-dimensional feature vectors of photopolymer 3D printing.

[0010] The collected historical printing data is labeled as normal data samples and fault data samples, and the fault prediction model is trained and optimized by combining high-dimensional feature vectors.

[0011] The training-based fault prediction model detects the real-time collected printing parameters, compares the prediction results with the preset warning threshold, and issues a warning based on the fault classification result if the threshold is exceeded.

[0012] As a further aspect of the present invention, the photopolymerization 3D printing equipment is equipped with a resin tank temperature sensor, a light source intensity sensor, a vibration sensor, a position sensor, and a current sensor, which are used to collect printing parameters during photopolymerization 3D printing: resin temperature, ultraviolet light intensity, vibration frequency, vibration amplitude, and Z-axis height.

[0013] As a further aspect of the present invention, when preprocessing the collected printing parameters, the collected printing parameters are subjected to moving average processing to determine the window size of the moving average; based on the data sampling frequency and the set smoothing degree, the average value of the printing parameters at each time point before and after that time is calculated; smoothed time series data are obtained, including smoothed resin temperature, ultraviolet light intensity, vibration data within the set time period, and Z-axis height position signal.

[0014] As a further aspect of the present invention, when preprocessing the collected printing parameters, the method further includes eliminating spike noise by median filtering of the smoothed time series data. By determining the window size of the median filter, the median value is taken for the data at each time point and the preceding and following moving windows, thereby removing sudden abnormal values ​​in the resin temperature, filtering out abnormal vibration peaks, and removing occasional current spikes.

[0015] As a further aspect of the present invention, when preprocessing the collected printing parameters, the method further includes identifying and processing outliers in the data after median filtering. By calculating the mean and standard deviation of the data corresponding to each sensor, the resin temperature, ultraviolet light intensity, vibration frequency and amplitude, and Z-axis height are Z-score standardized to obtain the preprocessed data.

[0016] As a further aspect of the present invention, when extracting multidimensional features from photopolymer 3D printing, the smoothed time series data in the preprocessed data is segmented according to the time series, and the temporal features of each sensor data in each time period are calculated. The extracted temporal features include the mean, standard deviation, the maximum and minimum values ​​of extreme value changes in each time period, and the skewness and kurtosis of the data distribution. The time series data is converted to the frequency domain, and spatial features are extracted. The spatial features include the frequency domain features extracted by Fourier transform and the power spectral density at the same frequency.

[0017] As a further aspect of the present invention, the AI-driven photopolymerization 3D printing fault prediction method further includes spatial feature extraction based on graph neural networks, comprising the following steps:

[0018] Collect sensor data and physical structure data of the photopolymer 3D printing equipment;

[0019] The physical structure of the photopolymer 3D printer is modeled as a graph, where nodes represent sensors and edges represent data flow.

[0020] Sensor data is mapped to the corresponding graph nodes to form node feature vectors, where each node's feature vector contains the multidimensional data collected by that node;

[0021] Based on graph convolutional networks as graph neural network architecture, a graph neural network model is constructed, in which the input layer receives node features, the hidden layer performs feature aggregation and transformation, and the output layer generates the final feature representation.

[0022] The preprocessed graph data is divided into a training set, a validation set, and a test set. The graph neural network model is trained using the training set, and the trained graph neural network model is used to validate the graph data in the test set. High-dimensional feature vectors of each node are extracted.

[0023] The extracted spatial features are fused with temporal and frequency domain features to form a comprehensive feature vector, which is then used to train the fault prediction model.

[0024] As a further aspect of the present invention, temporal and spatial feature analysis is performed to obtain a high-dimensional feature vector for photopolymerization 3D printing, including the following steps: aligning data from different sensors according to timestamps, and concatenating the extracted temporal and spatial features into a high-dimensional feature vector.

[0025] As a further aspect of the present invention, the training and optimization of the constructed fault prediction model includes the following steps:

[0026] Collect historical printing data generated during the photopolymerization 3D printing process, including temperature, ultraviolet light intensity, vibration frequency, vibration amplitude, and Z-axis height;

[0027] Historical printed data is divided into normal data samples and faulty data samples based on whether the data is abnormal or faulty within a time period, and then randomly divided into training set and test set according to a certain ratio.

[0028] The selected model is trained using high-dimensional feature vectors from the training set and corresponding normal or fault labels. The model's hyperparameters are then adjusted, and cross-validation is performed using a test set to obtain the trained and optimized fault prediction model.

[0029] In a second aspect, the present invention also includes a computer device comprising: at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the AI-driven photopolymerization 3D printing fault prediction method.

[0030] Thirdly, the present invention also includes a computer-readable storage medium storing computer instructions for causing the computer to execute the AI-driven photopolymerization 3D printing fault prediction method.

[0031] Compared with existing technologies, the AI-driven photopolymerization 3D printing fault prediction method provided by this invention, through real-time data acquisition, comprehensive feature extraction, and advanced machine learning models, achieves fault prediction and early warning in the 3D printing process, and has the following beneficial effects:

[0032] 1. Improved printing quality and success rate, enabling real-time monitoring and dynamic adjustment.

[0033] The photopolymer 3D printing fault prediction method of the present invention monitors various parameters of the printer in real time (such as temperature, humidity, light intensity, material flow, etc.) through sensors. Real-time monitoring can detect abnormalities in a timely manner, dynamically adjust printing parameters based on real-time data, optimize the printing process, avoid quality problems, and improve the printing success rate.

[0034] 2. It enables preventative maintenance and fault early warning, facilitating preventative maintenance and reducing downtime.

[0035] This invention helps to prevent equipment damage and extend equipment life by predicting failures in advance and performing timely maintenance and upkeep; it also analyzes equipment operating data to identify potential failure modes, reduce unexpected downtime, and improve equipment utilization and production efficiency.

[0036] 3. Improved production efficiency and enabled intelligent early warning.

[0037] By monitoring and dynamically adjusting printing parameters in real time, the printing process is optimized, production efficiency is improved, reliance on manual monitoring and intervention is reduced, and the level of production automation is enhanced. Multi-dimensional data is collected through various sensors, and preprocessed and feature extracted to fully utilize the data and improve prediction accuracy. By combining temporal features, frequency domain features, and spatial features, various data in the printing process are comprehensively analyzed to improve the accuracy of fault detection. It is applicable to different types of photopolymer 3D printing equipment and various printing materials, and has broad application prospects.

[0038] In summary, the AI-driven photopolymerization 3D printing fault prediction method of the present invention can not only significantly improve printing quality and efficiency, but also effectively reduce costs, extend equipment life, and provide users with a better user experience and higher production capacity, thus having significant technical and economic benefits.

[0039] These or other aspects of the invention will become more apparent from the following description of embodiments. It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the accompanying drawings used in the description of the exemplary embodiments or related technologies will be briefly introduced below. The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation thereof. In the drawings:

[0041] Figure 1 This is a flowchart of an AI-driven photopolymerization 3D printing fault prediction method according to an embodiment of the present invention.

[0042] Figure 2 This is a flowchart illustrating spatial feature extraction based on graph neural networks in the AI-driven photopolymerization 3D printing fault prediction method of this invention.

[0043] Figure 3 This is a flowchart illustrating the training and optimization of the fault prediction model constructed in the AI-driven photopolymerization 3D printing fault prediction method of this invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0046] The technical solutions in the exemplary embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described exemplary embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] The technical solution of the present invention will be further described below with reference to specific embodiments:

[0048] See Figure 1 As shown, Figure 1 A flowchart illustrating an AI-driven photopolymerization 3D printing fault prediction method provided by the present invention. One embodiment of the present invention provides an AI-driven photopolymerization 3D printing fault prediction method, comprising the following steps:

[0049] Step S10: Based on the sensors installed on the photopolymerization 3D printing equipment, the printing parameters are collected in real time, and the collected printing parameters are preprocessed by the data processing module to eliminate noise and outliers.

[0050] In this step, the photopolymer 3D printing equipment is equipped with a resin tank temperature sensor, a light source intensity sensor, a vibration sensor, a position sensor, and a current sensor, which are used to collect printing parameters during photopolymer 3D printing: resin temperature, ultraviolet light intensity, vibration frequency, vibration amplitude, and Z-axis height.

[0051] In this embodiment, when preprocessing the collected printing parameters, the collected printing parameters are subjected to moving average processing to determine the window size of the moving average; based on the data sampling frequency and the set smoothing degree, the average value of the printing parameters at each time point before and after that time is calculated; smoothed time series data are obtained, including smoothed resin temperature, ultraviolet light intensity, vibration data within a set time period, and Z-axis height position signal.

[0052] The data preprocessing of the collected printing parameters includes eliminating spike noise by median filtering of smooth time series data. By determining the window size of the median filter, the median value of the data at each time point and before and after the moving window is taken to remove sudden abnormal values ​​in resin temperature, filter out abnormal vibration peaks, and remove occasional current spikes.

[0053] The data preprocessing of the collected printing parameters includes identifying and processing outliers in the median-filtered data. By calculating the mean and standard deviation of the data corresponding to each sensor, the resin temperature, ultraviolet light intensity, vibration frequency and amplitude, and Z-axis height are Z-score standardized to obtain the preprocessed data.

[0054] In the above steps, the AI-driven photopolymerization 3D printing fault prediction method collects data through multiple sensors and uses data preprocessing technology to eliminate noise and outliers, ensuring high accuracy of the acquired printing parameter data and providing a reliable data foundation for subsequent fault prediction.

[0055] The moving average formula used in the moving average processing is as follows:

[0056]

[0057] In the formula, This is the moving average over time t; The window size for the determined moving average; Then it is the first Data collected at each time point.

[0058] When using median filtering to eliminate spike noise, the formula for calculating the median filter value at time point t is:

[0059]

[0060] in, This is the median filtered value at time point t; The determined window size for median filtering; In the first Data collected at each time point.

[0061] For example, suppose that during a photopolymer 3D printing process, the sensor collects the following data (partial data per unit time):

[0062] Resin temperature (°C): 30, 31, 32, 35, 30, 29, 28, 27, 30, 29;

[0063] Ultraviolet light intensity (mW / cm²): 10, 11, 10, 12, 11, 10, 9, 8, 10, 11;

[0064] Vibration frequencies (Hz): 50, 51, 50, 52, 53, 54, 55, 56, 55, 54;

[0065] Vibration amplitude (mm): 0.5, 0.6, 0.7, 0.8, 0.7, 0.6, 0.5, 0.4, 0.5, 0.6;

[0066] Z-axis height (mm): 10, 10.1, 10.2, 10.3, 10.2, 10.1, 10, 9.9, 10, 10.1;

[0067] Then, data preprocessing is performed, starting with a moving average, where:

[0068] Assuming window size =3, then the moving average of the resin temperature at the 5th time point is:

[0069]

[0070] Then, median filtering is performed, where:

[0071] Assuming window size =3, then the median filtered value of the ultraviolet light intensity at the 5th time point is:

[0072]

[0073] Finally, Z-score standardization is performed, where:

[0074] Assuming the mean resin temperature μ = 30℃ and the standard deviation σ = 2℃, the Z-score standardized value of the resin temperature at the 5th time point is:

[0075]

[0076] In this embodiment, by employing steps such as sensor data acquisition, moving average processing, median filtering, and Z-score standardization, noise and outliers can be effectively eliminated, improving data quality. The preprocessed data provides a reliable input foundation for the AI ​​system, thereby improving the accuracy and timeliness of photopolymer 3D printing fault prediction, ultimately enhancing print quality and equipment operational stability.

[0077] Step S20: Extract multidimensional features of photopolymer 3D printing from the preprocessed data, and perform temporal and spatial feature analysis to obtain high-dimensional feature vectors of photopolymer 3D printing.

[0078] In this step, when extracting the multidimensional features of photopolymer 3D printing, the smoothed time series data in the preprocessed data is segmented according to the time series, and the time series features of each sensor data in each time period are calculated. The extracted time series features include the mean, standard deviation, the maximum and minimum values ​​of extreme value changes in each time period, and the skewness and kurtosis of the data distribution. The time series data is then converted to the frequency domain to extract spatial features, which include the frequency domain features extracted by Fourier transform and the power spectral density at the same frequency.

[0079] In this embodiment, temporal and spatial feature analysis is performed to obtain a high-dimensional feature vector for photopolymerization 3D printing. The steps include: aligning data from different sensors according to timestamps, and concatenating the extracted temporal and spatial features into a high-dimensional feature vector.

[0080] Among them, see Figure 2 As shown, the AI-driven photopolymerization 3D printing fault prediction method also includes spatial feature extraction based on graph neural networks, comprising the following steps:

[0081] Step S201: Collect sensor data and physical structure data of the photopolymer 3D printing equipment;

[0082] Step S202: Model the physical structure of the photopolymer 3D printer as a graph, where nodes represent sensors and edges represent data flow;

[0083] Step S203: Map the sensor data to the corresponding graph nodes to form node feature vectors, wherein the feature vector of each node contains the multidimensional data collected by that node;

[0084] Step S204: Based on the graph convolutional network as the graph neural network architecture, construct a graph neural network model, wherein the input layer is defined to receive node features, the hidden layer performs feature aggregation and transformation, and the output layer generates the final feature representation;

[0085] Step S205: Divide the preprocessed graph data into a training set, a validation set, and a test set. Use the training set to train the graph neural network model, and apply the trained graph neural network model to validate the graph data in the test set. Extract the high-dimensional feature vectors of each node.

[0086] Step S206: The extracted spatial features are fused with temporal and frequency domain features to form a comprehensive feature vector, and the fused comprehensive feature vector is used to train the fault prediction model.

[0087] In this embodiment, the AI-driven photopolymerization 3D printing fault prediction method extracts temporal and spatial features from the preprocessed data, further constructs a high-dimensional feature vector, and extracts spatial features based on a graph neural network (GNN). Finally, the temporal, spatial, and frequency domain features are fused to form a comprehensive feature vector, which is used for training and validation of the fault prediction model.

[0088] For example, when obtaining the high-dimensional feature vector of photopolymer 3D printing, assume that the following sensor data were collected during the photopolymer 3D printing process:

[0089] Resin temperature: 30℃, 31℃, 32℃, 33℃, 31℃, 30℃;

[0090] Vibration frequencies: 50Hz, 51Hz, 50Hz, 52Hz, 51Hz, 50Hz.

[0091] Therefore, when performing time-series feature extraction, the average value and standard deviation of the resin temperature are calculated, where:

[0092] The average resin temperature Mean = (30 + 31 + 32 + 33 + 31 + 30) ÷ 6 ≈ 31℃;

[0093] Standard deviation of resin temperature Std= .

[0094] Then, the time-series features, frequency-domain features, and graph neural network-based feature vectors are concatenated to form a comprehensive feature vector for fault prediction.

[0095] By combining temporal, frequency, and spatial features, and especially by introducing spatial structure modeling using graph neural networks, this AI-driven photopolymerization 3D printing fault prediction method can accurately capture multidimensional features during equipment operation, improving the ability to predict potential faults. Through these complex feature extraction and fusion processes, the resulting high-dimensional comprehensive feature vector provides richer information for the fault prediction model, improving its accuracy, robustness, and timeliness, and ensuring the stability and reliability of the printing process.

[0096] Step S30: Based on the collected historical printing data, label them as normal data samples and fault data samples, and train and optimize the constructed fault prediction model by combining high-dimensional feature vectors.

[0097] In this step, see Figure 3 As shown, the training and optimization of the constructed fault prediction model includes the following steps:

[0098] Step S301: Collect historical printing data generated during the photopolymerization 3D printing process, including temperature, ultraviolet light intensity, vibration frequency, vibration amplitude, and Z-axis height;

[0099] Step S302: Divide the historical printed data into normal data samples and faulty data samples according to whether the data is abnormal or faulty within the time period, and randomly divide them into training set and test set according to the proportion.

[0100] Step S303: Train the selected model using the high-dimensional feature vectors in the training set and the corresponding normal or fault labels, adjust the model's hyperparameters, and perform cross-validation using the test set to obtain the trained and optimized fault prediction model.

[0101] This step involves collecting historical data generated by various sensors during the photopolymerization 3D printing process, covering the following key data:

[0102] Temperature: Temperature data inside the resin or equipment.

[0103] Ultraviolet light intensity: The intensity of ultraviolet light irradiation used in the curing process.

[0104] Vibration frequency and vibration amplitude: Vibration data during equipment operation.

[0105] Z-axis height: The Z-axis movement of the equipment.

[0106] During data annotation, data is labeled as normal samples and fault samples based on anomalies observed in the printing process in historical data. Fault criteria can include equipment downtime, printing failure, or parameters exceeding safe ranges. Data is randomly divided into training and test sets proportionally (80% for training, 20% for testing). During model training, a Support Vector Machine (SVM) classifier is used to optimize the objective, and a Random Forest is used to adjust the model's hyperparameters. Accuracy is used as the evaluation metric.

[0107] Assume the historical printing data of the photopolymer 3D printer is as follows:

[0108] Temperature: [30℃, 32℃, 33℃, 34℃, 32℃]

[0109] Ultraviolet light intensity: [100W / m2, 105W / m2, 108W / m2, 102W / m2, 110W / m2]

[0110] Vibration frequencies: [50Hz, 51Hz, 52Hz, 50Hz, 53Hz]

[0111] These historical print data are labeled as normal data or fault data depending on whether a fault occurred.

[0112] Then, by combining temporal features (mean and standard deviation of temperature), frequency domain features (power spectral density of vibration frequency), and spatial features, a comprehensive high-dimensional feature vector is formed:

[0113] Feature vector 1: [31.5, 2.5, 105, 50.5, ...];

[0114] Feature vector 2: [32, 1.8, 108, 51, ...];

[0115] These high-dimensional feature vectors, along with their corresponding labels (normal or faulty), are used to train the model and optimized through cross-validation. In photopolymer 3D printing fault prediction, data preprocessing and high-dimensional feature extraction provide the model with rich information. By using classification models such as SVM, random forest, and GBDT, combined with appropriate hyperparameter tuning and cross-validation, the model can effectively identify potential faults and improve the reliability and stability of the printing equipment.

[0116] Step S40: The fault prediction model based on training detects the real-time collected printing parameters, compares the obtained prediction results with the preset warning threshold, and issues a warning based on the fault classification result if the threshold is exceeded.

[0117] In this step, real-time collected data is input into the trained fault prediction model. Through the learned features and patterns, a predicted value is generated that can represent whether the equipment is in a normal or faulty state, serving as the probability of a certain fault occurring. A warning threshold is set as a critical point for judging "normal" or "fault". For example, if the predicted fault probability value P(fault|X) exceeds a certain set threshold T, it indicates that the equipment is at risk of failure. If the fault probability output by the model exceeds the threshold, the warning mechanism is triggered. For example, if the predicted value exceeds the minor fault threshold but is lower than the severe fault threshold, it can be judged as a moderate fault, and a warning of the corresponding level is issued.

[0118] For example, based on the magnitude of the predicted value, multiple levels of early warning can be set, where different fault classification thresholds are defined: , , .

[0119] Minor fault: ;

[0120] Moderate fault: ;

[0121] Critical fault: .

[0122] For example, suppose that during the photopolymer 3D printing process, the system collects the following real-time data:

[0123] Temperature: 40℃;

[0124] Ultraviolet light intensity: 120W / m²;

[0125] Vibration frequency: 55Hz;

[0126] Z-axis height: 0.4mm.

[0127] The data is input into a trained random forest fault prediction model, and the model outputs a fault prediction probability of 0.8.

[0128] If the system's pre-warning threshold is set to T=0.75, and the predicted value exceeds the preset threshold, the system determines it to be a moderate fault and issues a pre-warning prompt to the operator to check the equipment.

[0129] If the fault prediction value is 0.90, it exceeds the threshold for a severe fault. If the value is 0.85, the system will immediately shut down and issue an alarm, requiring the equipment to be inspected and repaired.

[0130] This invention presents an AI-driven photopolymerization 3D printing fault prediction method that combines historical data with real-time acquired data, utilizing a trained model to monitor the equipment's operating status in real time. By comparing data with warning thresholds, the system can issue early warnings before faults occur, thereby reducing the risk of printing failures. Fault grading and automatic response mechanisms make the system more intelligent and can provide effective solutions for different fault levels, greatly improving equipment reliability and production efficiency.

[0131] It should be understood that although the above description follows a certain order, these steps are not necessarily executed in that order. Unless otherwise expressly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, some steps in this embodiment may include multiple steps or multiple stages, which are not necessarily completed at the same time, but may be executed at different times. The execution order of these steps or stages is not necessarily sequential, but may be performed alternately or in turn with other steps or at least a portion of the steps or stages in other steps.

[0132] In one embodiment, a computer device is also provided in the present invention, including at least one processor and a memory communicatively connected to the at least one processor, the memory storing instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the steps of the AI-driven photopolymerization 3D printing fault prediction method.

[0133] In one embodiment, the present invention also provides a computer-readable storage medium storing computer instructions for causing the computer to perform the steps of the AI-driven photopolymerization 3D printing fault prediction method.

[0134] Those skilled in the art will understand that implementing all or part of the processes in the above embodiments can be accomplished by a computer program, characterized by computer instructions, instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Furthermore, any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory.

[0135] Non-volatile memory may include read-only memory, magnetic tape, floppy disk, flash memory, or optical storage. Volatile memory may include random access memory or external cache memory. By way of illustration and not limitation, RAM can take many forms, such as static random access memory or dynamic random access memory.

[0136] 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, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AI-driven method for predicting faults in photopolymerization 3D printing, characterized in that, The method includes the following steps: The printing parameters are collected in real time by sensors installed on the photopolymerization 3D printing equipment, and the collected printing parameters are preprocessed by the data processing module to eliminate noise and outliers. Multidimensional features of photopolymer 3D printing are extracted from the preprocessed data, and temporal and spatial features are analyzed to obtain high-dimensional feature vectors of photopolymer 3D printing. The collected historical printing data is labeled as normal data samples and fault data samples, and the fault prediction model is trained and optimized by combining high-dimensional feature vectors. The training-based fault prediction model detects the real-time collected printing parameters, compares the prediction results with the preset warning threshold, and issues a warning based on the fault classification result if the threshold is exceeded. It also includes spatial feature extraction based on graph neural networks, including the following steps: Collect sensor data and physical structure data of photopolymer 3D printing equipment; The physical structure of the photopolymer 3D printer is modeled as a graph, where nodes represent sensors and edges represent data flow. Sensor data is mapped to the corresponding graph nodes to form node feature vectors, where each node's feature vector contains the multidimensional data collected by that node; Based on graph convolutional networks as graph neural network architecture, a graph neural network model is constructed, in which the input layer receives node features, the hidden layer performs feature aggregation and transformation, and the output layer generates the final feature representation. The preprocessed graph data is divided into a training set, a validation set, and a test set. The graph neural network model is trained using the training set, and the trained graph neural network model is used to validate the graph data in the test set. High-dimensional feature vectors of each node are extracted. The extracted spatial features are fused with temporal and frequency domain features to form a comprehensive feature vector, which is then used to train the fault prediction model.

2. The AI-driven photopolymerization 3D printing fault prediction method as described in claim 1, characterized in that, The photopolymer 3D printing equipment is equipped with a resin tank temperature sensor, a light source intensity sensor, a vibration sensor, a position sensor, and a current sensor, which are used to collect printing parameters during photopolymer 3D printing: resin temperature, ultraviolet light intensity, vibration frequency, vibration amplitude, and Z-axis height.

3. The AI-driven photopolymerization 3D printing fault prediction method as described in claim 2, characterized in that, When preprocessing the collected printing parameters, the process also includes eliminating spike noise by median filtering of the smoothed time series data. By determining the window size of the median filter, the median value is taken for the data at each time point and before and after the moving window, thus removing sudden abnormal values ​​in the resin temperature, filtering out abnormal vibration peaks, and removing occasional current spikes.

4. The AI-driven photopolymerization 3D printing fault prediction method as described in claim 3, characterized in that, When preprocessing the collected printing parameters, it also includes identifying and processing outliers in the median-filtered data. By calculating the mean and standard deviation of the data corresponding to each sensor, the resin temperature, ultraviolet light intensity, vibration frequency and amplitude, and Z-axis height are Z-score standardized to obtain the preprocessed data.

5. The AI-driven photopolymerization 3D printing fault prediction method as described in claim 4, characterized in that, When extracting multidimensional features from photopolymer 3D printing, the smoothed time series data in the preprocessed data is segmented according to the time series, and the temporal features of each sensor data in each time period are calculated. The extracted temporal features include the mean, standard deviation, the maximum and minimum values ​​of extreme value changes in each time period, and the skewness and kurtosis of the data distribution. The time series data is then converted to the frequency domain to extract spatial features, including the frequency domain features extracted by Fourier transform and the power spectral density at the same frequency.

6. The AI-driven photopolymerization 3D printing fault prediction method as described in claim 1, characterized in that, The process of performing temporal and spatial feature analysis to obtain a high-dimensional feature vector for photopolymer 3D printing includes the following steps: aligning data from different sensors according to timestamps, and concatenating the extracted temporal and spatial features into a high-dimensional feature vector.

7. The AI-driven photopolymerization 3D printing fault prediction method as described in claim 6, characterized in that, The training and optimization of the constructed fault prediction model includes the following steps: Collect historical printing data generated during the photopolymerization 3D printing process, including temperature, ultraviolet light intensity, vibration frequency, vibration amplitude, and Z-axis height; Historical printed data is divided into normal data samples and faulty data samples based on whether the data is abnormal or faulty within a time period, and then randomly divided into training set and test set according to a certain ratio. The selected model is trained using high-dimensional feature vectors from the training set and corresponding normal or fault labels. The model's hyperparameters are then adjusted, and cross-validation is performed using a test set to obtain the trained and optimized fault prediction model.

8. A computer device, characterized in that, include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the AI-driven photopolymerization 3D printing fault prediction method according to any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the AI-driven photopolymerization 3D printing fault prediction method according to any one of claims 1-7.

Images