Printing failure detection method and system for 3D printer

Abstract

The invention relates to the technical field of 3D printers, in particular to a printing failure detection method and system for a 3D printer. The method comprises the following steps that fused deposition modeling abnormal state data are extracted by obtaining a 3D printer operation log, then an abnormal wire drawing mechanism is reversely deduced, and the edge warping probability between printing layers is estimated; then, carrying out cracking risk calculation on the edge warping probability data to obtain printing interlayer cracking risk data; and based on the convolutional neural network, constructing a printing interlayer cracking risk identification model, and sending the model to a 3D printer control center for executing printing failure detection. According to the method, the 3D printing failure detection technology is optimized, so that the 3D printing failure detection technology is more perfect.

Description

Technical Field

[0001] The present invention relates to the technical field of 3D printers, and in particular to a 3D printer printing failure detection method and system. Background Art

[0002] With the rapid development of 3D printing technology, 3D printers have been widely used in manufacturing, medical, aerospace, automotive and other fields. 3D printing, as a layer-by-layer additive manufacturing method, has a high degree of freedom and design flexibility, and can manufacture objects of complex shapes. However, in practical applications, various problems are often encountered in the 3D printing process, leading to printing failures, such as warping, cracking, falling off, defects between printing layers, etc. These problems not only affect the printing effect, but also lead to waste of raw materials, extended printing time, and even damage to printing equipment. Therefore, how to detect and predict abnormal conditions in the printing process in real time and discover potential printing problems in time has become the key to improving 3D printing quality, reducing production costs, and improving efficiency. In the FDM printing process, abnormal drawing is one of the key factors affecting printing quality. Abnormal wire drawing usually occurs after the nozzle extrude the material. Due to the uneven cooling of the hot-melt filamentary material in the molten state in the air, the filamentary material fails to deposit as expected, but instead appears filamentous stretching, accumulation or adhesion. This abnormal wire drawing not only affects the surface quality of the print, but also causes uneven adhesion between the printed layers, which further causes the warping problem. The warping between layers is due to the difference in shrinkage rate during the cooling process of the material. In particular, when the cooling rates of different layers are inconsistent, the interlayer adhesion is reduced, which leads to warping. Warping not only affects the dimensional accuracy of the printed part, but when the degree of warping is large or the bonding force between the material layers is insufficient, the different layers of the printed part cannot form a stable bond, which eventually leads to interlayer separation or obvious cracks, and finally causes the risk of cracking. However, a traditional 3D printer printing failure detection method has low risk identification accuracy for abnormal conditions (such as wire drawing, warping and cracking), and weak ability to trace the causes of abnormal conditions, resulting in large errors in printing failure detection. Summary of the invention

[0003] Based on this, it is necessary to provide a 3D printer printing failure detection method and system to solve at least one of the above technical problems.

[0004] To achieve the above object, a 3D printer printing failure detection method is provided, the method comprising the following steps: Step S1: obtaining a 3D printer operation log; performing fused deposition modeling abnormal state extraction on the 3D printer operation log to obtain fused deposition modeling abnormal state data; Wherein, step S2 comprises: Step S21: obtaining hot-melt filament material state data; Step S22: performing printing abnormal morphological deformation identification based on the fused deposition modeling abnormal state data to obtain printing abnormal morphological deformation data; Step S23: reversely inferring the abnormal wire drawing mechanism of the printed abnormal morphological deformation data according to the state data of the hot-melt filamentary material to obtain the abnormal wire drawing mechanism; Step S24: estimating the probability of warping between printing layers based on the printing abnormal morphological deformation data and the abnormal wire drawing mechanism, and obtaining the probability data of warping between printing layers; Step S25: performing cracking risk calculation on the probability data of warping between printed layers to obtain cracking risk data between printed layers; Step S3: constructing a printing layer crack risk identification model based on the printing layer crack risk data based on the convolutional neural network to obtain a printing layer crack risk identification model; Step S4: Sending the inter-layer crack risk identification model to the 3D printer control center to perform printing failure detection.

[0005] Preferably, step S1 comprises the following steps: Step S11: Obtaining the 3D printer operation log; Step S12: performing data cleaning on the 3D printer operation log to obtain a 3D printer operation cleaning log; Step S13: extracting abnormal state of fused deposition modeling from the cleaning log of the 3D printer to obtain abnormal state data of fused deposition modeling.

[0006] Preferably, step S23 includes the following steps: Step S231: performing a thermal viscosity drop gradient analysis on the state data of the hot-melt filamentary material to obtain thermal viscosity drop gradient data; Step S232: performing a thermoplastic elongation at break change simulation on the state data of the hot-melt filamentary material based on the thermal viscosity decrease gradient data to obtain thermoplastic elongation at break change data; Step S233: analyzing the abnormal deformation contour parameters of the printed abnormal morphological deformation data to obtain the abnormal deformation contour parameters; Step S234: performing local abnormal deformation thermoplastic elongation distribution inversion on the abnormal deformation contour parameters according to the thermoplastic elongation at break change data to obtain the abnormal deformation thermoplastic distribution elongation; Step S235: performing abnormal calculation of printing retraction distance / speed based on abnormal deformation thermoplastic distribution elongation and abnormal deformation profile parameters to obtain abnormal printing retraction distance / speed; Step S236: reversely infer the abnormal wire drawing mechanism of the abnormal deformation profile parameters according to the abnormal printing retraction distance / speed and the abnormal deformation thermoplastic distribution elongation to obtain the abnormal wire drawing mechanism.

[0007] Preferably, step S24 includes the following steps: Step S241: performing internal void distribution identification on the printed abnormal morphology deformation data to obtain abnormal morphology internal void distribution data; Step S242: reversely infer the change of hot melt tension of the material based on the abnormal wire drawing mechanism to obtain the change data of hot melt tension of the material; Step S243: Calculate the incremental gradient of the material hot melt cooling shrinkage rate according to the abnormal morphology internal void distribution data and the material hot melt tension change data to obtain the incremental gradient of the hot melt cooling shrinkage rate; Step S244: performing a multivariate linear regression analysis on the incremental gradient of the hot melt cooling shrinkage rate to obtain the incremental linear regression gradient of the shrinkage rate; Step S245: estimating the probability of warping between printing layers based on the linear regression gradient of the shrinkage increment to obtain the probability data of warping between printing layers.

[0008] Preferably, step S25 comprises the following steps: Step S251: evaluating the non-uniform cooling rate of the inter-layer warping probability data to obtain the inter-layer warping non-uniform cooling rate; Step S252: performing numerical fluctuation simulation of internal stress accumulation on the probability data of interlayer warping according to the non-uniform cooling rate of interlayer warping to obtain numerical fluctuation data of interlayer internal stress; Step S253: performing interlayer adhesion instability risk interval analysis based on interlayer internal stress numerical fluctuation data and interlayer warping non-uniform cooling rate to obtain an interlayer adhesion instability risk interval; Step S254: performing cracking risk calculation according to the interlayer adhesion instability risk interval to obtain interlayer cracking risk data.

[0009] Preferably, step S3 comprises the following steps: Step S31: performing convolution calculation on the risk level of cracking between printed layers to obtain convolution data of the risk of cracking between printed layers; Step S32: performing logic learning on the risk convolution data of cracking between printed layers to obtain logic data of cracking risk between printed layers; Step S33: constructing a printing layer crack risk identification model based on the printing layer crack risk logic data based on the convolutional neural network to obtain the printing layer crack risk identification model.

[0010] Preferably, step S33 includes the following steps: Step S331: extracting multi-scale features from the printed interlayer crack risk logic data to obtain interlayer crack multi-scale risk data; Step S332: dividing the interlayer cracking multi-scale risk data into a training set and a test set, and obtaining an interlayer cracking multi-scale risk training set and an interlayer cracking multi-scale risk test set respectively; Step S333: constructing an initial printing interlayer crack risk identification model for the interlayer crack multi-scale risk training set based on a convolutional neural network to obtain an initial printing interlayer crack risk identification model; Step S334: performing a model test on the initial printing interlayer crack risk identification model according to the interlayer crack multi-scale risk test set to obtain a printing interlayer crack risk identification model.

[0011] Preferably, the present invention further provides a 3D printer printing failure detection system, which is used to execute the 3D printer printing failure detection method as described above, and the 3D printer printing failure detection system comprises: The abnormal state extraction module is used to obtain the operation log of the 3D printer; perform fused deposition modeling abnormal state extraction on the operation log of the 3D printer to obtain fused deposition modeling abnormal state data; The cracking risk calculation module is used to reverse the abnormal wire drawing mechanism based on the abnormal state data of fused deposition modeling to obtain the abnormal wire drawing mechanism; estimate the probability of warping between printing layers based on the abnormal wire drawing mechanism to obtain the probability data of warping between printing layers; perform cracking risk calculation on the probability data of warping between printing layers to obtain the risk data of cracking between printing layers; A cracking risk identification model building module is used to build a printing interlayer cracking risk identification model based on the printing interlayer cracking risk data based on a convolutional neural network to obtain a printing interlayer cracking risk identification model; The detection execution module is used to send the printing layer crack risk identification model to the 3D printer control center to perform printing failure detection.

[0012] The beneficial effect of the present invention is that by obtaining the operation log of the 3D printer, the system can record various key parameters in the printing process in real time, such as temperature, printing speed, nozzle status, etc. These data provide valuable information resources for subsequent analysis. By extracting the abnormal state of fused deposition modeling from the operation log, abnormal conditions in the printing process, such as abnormal temperature fluctuations and nozzle blockage, can be accurately identified. This process not only helps to capture potential printing problems in a timely manner, but also provides an accurate data basis for subsequent mechanism analysis and risk prediction, avoiding omissions and delays in traditional manual monitoring. Based on the extracted abnormal state data of fused deposition modeling, the system reversely infers the abnormal wire drawing mechanism and analyzes the warping phenomenon between printing layers caused by uneven wire drawing, overheating of materials, etc. during the printing process. These abnormal wire drawing mechanisms reflect the root causes of physical deformation and defects in the printing process. Based on these mechanisms, the system further estimates the probability of warping between printing layers, and calculates the cracking risk through the cracking risk calculation model. This step can effectively predict structural problems in the printing process, provide a basis for subsequent measures, and thus improve the printing success rate. Through deep learning analysis of the interlayer cracking risk data using a convolutional neural network (CNN), the system can automatically identify the risk of interlayer cracking during the printing process. In this process, the CNN model extracts potential patterns and rules by learning a large amount of abnormal data, and then establishes an accurate cracking risk identification model. This model has a strong generalization ability and can accurately identify the risk of cracking under different printing conditions and environments. This step not only reduces the need for manual intervention, but also improves the accuracy and efficiency of risk identification, and timely detects problems and issues early warning. The constructed interlayer cracking risk identification model is sent to the 3D printer control center in real time. This operation enables the printer to obtain risk identification results in real time during the printing process and take corresponding preventive measures based on the risks predicted by the model. For example, the control center can automatically correct the problem that causes cracking by adjusting printing parameters (such as temperature, speed, path, etc.). This closed-loop control method greatly improves the adaptability and intelligence level of the printing process, making printing failure detection and risk management more efficient and accurate, greatly reducing the probability of printing failure, and ensuring the stability of the printing process and the quality of the final product. In the existing technology, the independent identification of wire drawing, edge warping and cracking has two key defects: lack of contextual association, inability to understand the evolution relationship between defects, and high false alarm rate: single feature recognition is easily affected by environmental interference and random factors. This case understands the complete evolution process from wire drawing to edge warping to cracking, effectively reducing the false alarm rate and improving recognition accuracy.Therefore, the present invention is an optimization processing of a traditional 3D printer printing failure detection method, which solves the problem that the traditional 3D printer printing failure detection method has low risk identification accuracy for abnormal conditions (such as wire drawing, warping and cracking), and weak ability to trace the causes of abnormal conditions, thereby causing large printing failure detection errors. The present invention improves the risk identification accuracy for abnormal conditions (such as wire drawing, warping and cracking), and the ability to trace the causes of abnormal conditions, thereby reducing the error of printing failure detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A schematic diagram of the steps of a 3D printer printing failure detection method; Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart; Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. DETAILED DESCRIPTION

[0014] See also Figures 1 to 3 , a 3D printer printing failure detection method, the method comprising the following steps: Step S1: obtaining a 3D printer operation log; performing fused deposition modeling abnormal state extraction on the 3D printer operation log to obtain fused deposition modeling abnormal state data; Step S2: reversely infer the abnormal wire drawing mechanism based on the abnormal state data of fused deposition modeling to obtain the abnormal wire drawing mechanism; estimate the probability of warping between printed layers based on the abnormal wire drawing mechanism to obtain the probability data of warping between printed layers; perform cracking risk calculation on the probability data of warping between printed layers to obtain the risk data of cracking between printed layers; Step S3: constructing a printing layer crack risk identification model based on the printing layer crack risk data based on the convolutional neural network to obtain a printing layer crack risk identification model; Step S4: Sending the inter-layer crack risk identification model to the 3D printer control center to perform printing failure detection.

[0015] In the embodiment of the present invention, reference Figure 1 The above is a schematic diagram of the steps of a 3D printer printing failure detection method of the present invention. In this example, the 3D printer printing failure detection method includes the following steps: Step S1: obtaining a 3D printer operation log; performing fused deposition modeling abnormal state extraction on the 3D printer operation log to obtain fused deposition modeling abnormal state data; In an embodiment of the present invention, it is first necessary to obtain the operation log data of the 3D printer. The operation log of the printer is generally recorded by multiple sensors and control systems, covering key data such as the temperature of the printer, the state of the extruder, the movement trajectory of the nozzle, the material deposition, the cooling rate, etc. The acquired log data is usually stored in a time series manner and contains abnormal alarm information when the device is running. The log data can be exported from the printer control system through a serial communication interface or a network communication protocol (such as MQTT, HTTPAPI) and stored in a local database or log analysis system for processing. When preprocessing the operation log data of the 3D printer, data cleaning and format conversion are required. The main purpose of data cleaning is to remove redundant data, fill in missing values, unify the data format, and filter out data fields related to abnormal states of fused deposition modeling. The cleaned data needs to be extracted for abnormal states. The specific method is to calculate the change trend of the data based on a time series anomaly detection algorithm, such as a sliding window statistical method or a method based on dynamic time warping (DTW). For example, in the fused deposition process, the temperature of the nozzle should usually be kept within the range of ±2°C of the set value. If the temperature deviation at a certain time point exceeds this range and lasts for a period of time, it is determined that a temperature anomaly has occurred in this time period. In addition, the printer's extrusion speed, interlayer bonding pressure and other parameters also need to be detected for abnormalities, such as identifying abnormal points in multidimensional sensor data through the Mahalanobis distance or local outlier factor (LOF) method, and finally obtaining the abnormal state data of fused deposition modeling.

[0016] Step S2: reversely infer the abnormal wire drawing mechanism based on the abnormal state data of fused deposition modeling to obtain the abnormal wire drawing mechanism; estimate the probability of warping between printed layers based on the abnormal wire drawing mechanism to obtain the probability data of warping between printed layers; perform cracking risk calculation on the probability data of warping between printed layers to obtain the risk data of cracking between printed layers; In the embodiment of the present invention, based on the abnormal state data of fused deposition modeling, it is necessary to further infer the abnormal wire drawing mechanism. Abnormal wire drawing mainly occurs between the extrusion end and the printing layer. Due to the different rheological properties of the material at different temperatures and extrusion speeds, abnormal wire drawing can be attributed to the viscosity change of the hot-melt filamentary material and the retraction behavior of the extruder under specific working conditions. First, by obtaining the state data of the hot-melt filamentary material, including its melt index (MFI), viscosity curve, tensile properties and other parameters, combined with the experimentally measured nozzle temperature change curve and extrusion rate fluctuation data, the viscosity decrease gradient of the material at different temperatures is analyzed. During the experimental measurement, a high-precision thermocouple is used to monitor the nozzle outlet temperature, and the material fluidity curve at different temperatures is recorded. Secondly, based on the thermal viscosity decrease gradient of the material, the finite element analysis (FEA) method is used to simulate the change of the thermoplastic elongation at break of the material, and the melt index increment of the material at a specific temperature is calculated. For example, if the MFI of a certain PLA material is 6 g / 10 min at 190°C and increases to 10 g / 10 min at 210°C, it can be inferred that the fluidity of the material increases sharply in the range of 190°C~210°C, which leads to an increase in the probability of material wiredrawing during printing. By analyzing the retraction distance and speed of the extruder, the partial differential equation (PDE) is used to describe the flow behavior of the material, and the model is corrected in combination with experimental data, and finally the abnormal wiredrawing mechanism is obtained. The abnormal wiredrawing mechanism directly affects the probability of warping between printed layers, so it is necessary to estimate the probability of warping between layers based on abnormal morphological deformation data and material hot melt tension change data. First, the internal void distribution of the abnormal morphological deformation data is analyzed, and the distribution characteristics of the voids between printed layers are extracted using X-ray CT scanning technology or voxel reconstruction method based on optical measurement. For example, in the printing experiment of a certain ABS material, if the interlayer bubble rate exceeds 1.5%, the probability of warping increases significantly. Based on the abnormal wire drawing mechanism, the change of hot melt tension formed by the material during the wire drawing process is calculated, and the incremental gradient of cooling shrinkage rate of the material at different drawing speeds is analyzed by finite element simulation or experimental measurement. For example, a high-precision displacement sensor is used to measure the shrinkage rate of the material after cooling, and the shrinkage rate change under different temperature conditions is recorded to obtain the incremental shrinkage rate curve. Based on these data, the incremental shrinkage rate gradient is fitted by multivariate linear regression analysis, and the probability of warping between printed layers is calculated using the Bayesian probability model to obtain the interlayer warping probability data. When calculating the cracking risk of the interlayer warping probability data, it is necessary to combine the non-uniform cooling rate of the material and the interlayer stress accumulation for calculation. First, a thermal imager or thermocouple array is used to monitor the cooling process of the printed part, and the interlayer cooling rate distribution data is obtained. For example, when printing ABS material, if the interlayer cooling rate is in the range of 3℃ / s~5℃ / s, the interlayer stress is low, and if the cooling rate exceeds 10℃ / s, the interlayer stress accumulation increases significantly.Based on the cooling rate data, the numerical simulation method is used to calculate the numerical fluctuation of the internal stress between layers, and the risk interval of interlayer adhesion instability is identified through the distribution of monotonically increasing mutation points. The Newton iteration method is used to numerically solve the internal stress mutation points, and combined with the thermal stress analysis model, the risk data of cracking between printed layers is obtained. It needs to be explained that the main reason why abnormal wiredrawing can cause warping and cracking is the hot melt rheological properties and cooling shrinkage behavior of the material. In the fused deposition modeling process, hot-melt filamentary material will be extruded first. If the extrusion temperature fluctuates or the extrusion rate is unstable, the hot melt viscosity of the material will decrease, the printing retraction will be abnormal, or the cooling rate will be uneven, and then thin wiredrawing will be formed on the printing path. These wiredrawings not only affect the uniform deposition of the interlayer material, but also form non-uniform thermal shrinkage in local areas. Since the material will shrink in volume during the cooling process, the non-uniform deposition caused by abnormal wiredrawing will cause changes in the local cooling rate, and then lead to uneven distribution of internal stress between layers. When the internal stress accumulation exceeds the interlayer adhesion strength of the material, warping will occur, especially in a high cooling rate environment, the thermal shrinkage stress increases, making the warping trend more obvious. At the same time, excessive local interlayer stress will form microcracks, which will further expand in the subsequent printing process, and eventually cause interlayer cracking.

[0017] Step S3: constructing a printing layer crack risk identification model based on the printing layer crack risk data based on the convolutional neural network to obtain a printing layer crack risk identification model; In an embodiment of the present invention, a convolutional neural network (CNN) is used to construct a risk identification model for interlayer cracking based on the risk data of interlayer cracking. First, multi-scale feature extraction is performed on the risk data of interlayer cracking, including local stress distribution, material shrinkage gradient, printing path influencing factors, etc. The data is preprocessed using sliding window technology, and a suitable convolution kernel size (such as 3×3 or 5×5) is selected to extract risk features of different scales. Then, the data is divided into a training set and a test set, and the model performance is evaluated using a cross-validation method. During the training process, the Adam optimizer is used to adjust the learning rate, and the model parameters are optimized according to the loss function (such as mean square error or cross entropy loss). Finally, the generalization ability of the model is verified on the test set, and the hyperparameters are adjusted to obtain the final risk identification model for interlayer cracking.

[0018] Step S4: Sending the inter-layer crack risk identification model to the 3D printer control center to perform printing failure detection.

[0019] In an embodiment of the present invention, the trained inter-layer cracking risk identification model is deployed to the 3D printer control center and integrated with the real-time monitoring system of the printer. The operating data of the printer is collected in real time and input into the cracking risk identification model to calculate the failure probability under the current printing state. If the cracking risk is detected to exceed a threshold (such as 90%), the printer control system can trigger an early warning and take measures such as adjusting printing parameters, reducing printing speed or increasing extrusion temperature to reduce the risk of printing failure.

[0020] Step S1 includes the following steps: Step S11: Obtaining the 3D printer operation log; Step S12: performing data cleaning on the 3D printer operation log to obtain a 3D printer operation cleaning log; Step S13: extracting abnormal state of fused deposition modeling from the cleaning log of the 3D printer to obtain abnormal state data of fused deposition modeling.

[0021] In an embodiment of the present invention, the acquisition process of the 3D printer operation log includes multiple key links such as data collection, storage and time synchronization. The control system of the printer records the operation status of the device in real time through built-in sensors and control modules, and stores relevant information in the form of structured log files. The operation log usually includes parameters such as nozzle temperature, extruder status, printing speed, platform temperature, Z-axis height, material flow rate, cooling fan speed, etc., and these parameters are all timestamped. The data acquisition method mainly extracts data from the printer's firmware through a serial communication interface (such as UART, RS-232) or a network communication protocol (such as MQTT, HTTPAPI), and stores it in a local database or log file. The data storage format generally adopts CSV, JSON or binary format for subsequent data processing. When obtaining the log, it is necessary to ensure that the log time is synchronized for subsequent abnormal analysis. For example, using NTP (Network Time Protocol) or GPS time synchronization technology to ensure that all log data is recorded according to a unified time base to avoid abnormal detection errors caused by time offset. The obtained operation log usually contains a large amount of redundant data, missing values ​​and inconsistent formats, so the data needs to be cleaned to ensure the accuracy of subsequent analysis. First, the format of the log data is standardized, and the field names, data types and units are unified, such as the temperature unit is unified as degrees Celsius (℃), and the time format is converted to the ISO 8601 standard format. Secondly, the interpolation method is used to fill the missing data. For the parameter missing problem caused by communication loss during the printing process, linear interpolation, spline interpolation or local weighted regression methods are used to reconstruct the data. For example, if the nozzle temperature data at a certain time point is missing, linear interpolation calculation can be performed based on the temperature values ​​at the previous and next time points. Then, abnormal data is eliminated, and the box plot analysis method (IQR) or the 3σ principle is used to eliminate abnormal values. For example, when the number of pulses of the extruder stepper motor fluctuates by more than ±5σ of the average value in a short period of time, the data point can be regarded as abnormal data and deleted. Finally, data denoising is performed, and sliding mean filtering or Kalman filtering is used to remove sensor noise. For example, sliding mean calculation is performed using data points with a window size of 5 to smooth the temperature change curve. The cleaned log data is stored in a new data table or file for subsequent abnormal state extraction of fused deposition modeling. The process of extracting abnormal states of fused deposition modeling from the operation log involves steps such as multidimensional feature analysis, time series anomaly detection, and pattern matching. First, extract key process parameters, including nozzle temperature, extrusion flow rate, printing speed, cooling fan speed, platform temperature, and interlayer cooling time, and calculate their first-order and second-order derivatives to obtain parameter change trends. For example, calculate the time derivative of the nozzle temperature to determine whether the temperature change rate is abnormal.Secondly, a time series anomaly detection method is used, such as a time series comparison algorithm based on dynamic time warping (DTW), to perform a similarity analysis between the current printing data and the normal printing data to detect whether there is an abnormal pattern. For example, if the extrusion flow rate of a certain print deviates from the normal printing data by more than 15%, it can be determined that there is an abnormal molten deposition state in this time period. In addition, statistical methods are used to analyze the correlation between the parameters, such as calculating the Pearson correlation coefficient between the extrusion flow rate and the nozzle temperature to determine whether there is an abnormal mismatch. For example, under normal circumstances, the correlation coefficient between the nozzle temperature and the extrusion flow rate should be above 0.8. If it is lower than 0.6, there is a molten deposition anomaly. Finally, a threshold-based method is used to classify abnormal states, such as classifying temperature anomalies, extrusion anomalies, and interlayer cooling anomalies separately, and storing them in the abnormal state data table for subsequent abnormal analysis and printing failure detection.

[0022] Step S2 includes the following steps: Step S21: obtaining hot-melt filament material state data; Step S22: performing printing abnormal morphological deformation identification based on the fused deposition modeling abnormal state data to obtain printing abnormal morphological deformation data; Step S23: reversely inferring the abnormal wire drawing mechanism of the printed abnormal morphological deformation data according to the state data of the hot-melt filamentary material to obtain the abnormal wire drawing mechanism; Step S24: estimating the probability of warping between printing layers based on the printing abnormal morphological deformation data and the abnormal wire drawing mechanism, and obtaining the probability data of warping between printing layers; Step S25: performing cracking risk calculation on the inter-printing layer warping probability data to obtain inter-printing layer cracking risk data.

[0023] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes: Step S21: obtaining hot-melt filament material state data; In an embodiment of the present invention, the acquisition of state data of hot-melt filamentary materials involves real-time monitoring of the physical and thermodynamic properties of the materials, including key parameters such as the melting temperature, viscosity, tensile strength, cooling shrinkage, and extrusion rate of the materials. During data acquisition, the nozzle temperature is measured by a thermocouple sensor, and the number of pulses of the extruder stepper motor is recorded in combination with an optical encoder to calculate the actual material extrusion rate. In addition, a high-precision strain gauge is used to monitor the tension of the material during stretching, and an infrared thermal imager is used to measure the temperature distribution of the material after it is extruded from the nozzle. In the experiment, PLA (polylactic acid) was selected as the test material, and the viscosity change curve of the material was recorded at different temperatures (190°C, 200°C, 210°C, 220°C) and extrusion rates (5mm / s, 10mm / s, 15mm / s), and the relationship between tensile strength and temperature was calculated. The acquired data is stored in a time series database format for subsequent analysis.

[0024] Step S22: performing printing abnormal morphological deformation identification based on the fused deposition modeling abnormal state data to obtain printing abnormal morphological deformation data; In an embodiment of the present invention, in the process of identifying abnormal morphological deformation of printing based on abnormal state data of fused deposition modeling, the interlayer deposition trajectory is first analyzed to extract key parameters such as contour deviation, deposition width, and layer height uniformity of each layer. The printing process is photographed by a high-speed camera, and the abnormal morphology on the printing path is identified using an edge detection algorithm. The Canny edge detection method is used to extract the contour of the molten filament and calculate its deviation from the theoretical path. For areas with large deviations, the morphological analysis method is further used to calculate the area, length, and curvature distribution of the abnormal area. For example, if a deviation of more than 0.3mm is detected in the deposition trajectory of a certain layer at a printing speed of 200mm / s, it can be determined that the layer has abnormal morphological deformation. After the abnormal data is stored, statistical analysis is performed to determine whether it is a systematic abnormality, such as insufficient extrusion, excessive retraction, and overcooling shrinkage.

[0025] Step S23: reversely inferring the abnormal wire drawing mechanism of the printed abnormal morphological deformation data according to the state data of the hot-melt filamentary material to obtain the abnormal wire drawing mechanism; In the embodiment of the present invention, the abnormal wire drawing mechanism is reversed based on the thermoplastic characteristics and rheological characteristics of the material, and the material state data and the printing abnormal morphological deformation data are used to deduce the mechanism of wire drawing. First, the thermal viscosity drop gradient of the hot-melt filamentary material is calculated, the Arrhenius equation is used to describe the viscosity change trend of the material, and the temperature-viscosity curve of PLA is fitted in combination with the experimental data. Secondly, the deformation behavior of the material is simulated based on the change of thermoplastic elongation at break, and the tensile strength and elongation at break of the material at different temperatures are calculated using the finite element analysis (FEA) method. For example, at 190°C, the elongation at break of PLA is 3.2%, which increases to 5.8% at 220°C. Then, the profile parameters of abnormal deformation during printing, such as wire drawing length and diameter changes, are analyzed, and the thermoplastic elongation distribution of local abnormal deformation is calculated. Combined with the thermoplastic characteristics of the material, the abnormal printing retraction distance and speed are calculated based on the Newton-Raphson method, thereby deducing the formation mechanism of abnormal wire drawing.

[0026] Step S24: estimating the probability of warping between printing layers based on the printing abnormal morphological deformation data and the abnormal wire drawing mechanism, and obtaining the probability data of warping between printing layers; In an embodiment of the present invention, the probability estimation of warping between printed layers is based on abnormal morphological deformation data and abnormal wire drawing mechanism, and the probability evaluation is performed by calculating the internal void distribution, material hot melt tension change and cooling shrinkage rate increment. First, the distribution of voids inside the print is analyzed by X-ray CT scanning technology, and the influence of void rate on the interlayer bonding strength is calculated. Secondly, based on the material hot melt tension model, the relationship curve of hot melt tension with temperature is fitted using tensile test data, and the hot melt tension change under different printing conditions is calculated in combination with the abnormal wire drawing mechanism. For example, at 220°C, the hot melt tension of PLA drops to 0.12MPa, while at 200°C it is 0.25MPa. Finally, combined with the void rate, hot melt tension change and cooling shrinkage rate increment, the probability of warping between layers is calculated by multivariate linear regression analysis, and the analysis results are stored.

[0027] Step S25: performing cracking risk calculation on the inter-printing layer warping probability data to obtain inter-printing layer cracking risk data.

[0028] In an embodiment of the present invention, the calculation of the risk of cracking between printed layers is based on the probability data of interlayer warping, and the internal stress accumulation, cooling rate distribution and instability risk of interlayer adhesion are comprehensively evaluated. First, the interlayer internal stress distribution is calculated using the finite element method, and a heat conduction and stress coupling model is established to simulate the stress accumulation caused by the temperature gradient during the cooling process of the material. Secondly, the non-uniform cooling rate of interlayer warping is evaluated, the interlayer temperature distribution is analyzed based on the thermal imager data, and the cooling rate difference between different layers is calculated using the Fourier heat conduction equation. For example, at a platform temperature of 60°C, the cooling rate of the top layer reaches -15°C / s, while that of the middle layer is only -5°C / s, indicating that there is a large temperature difference, resulting in instability of interlayer stress. Finally, based on the internal stress numerical fluctuation data and the change in cooling rate, the adhesion instability risk interval analysis is performed, the interlayer stress mutation point is calculated using the Newton iteration method, and the cracking risk is numerically calculated. The calculated cracking risk data is used for the training of subsequent risk identification models to improve the accuracy of printing failure detection.

[0029] Step S23 includes the following steps: Step S231: performing a thermal viscosity drop gradient analysis on the state data of the hot-melt filamentary material to obtain thermal viscosity drop gradient data; Step S232: performing a thermoplastic elongation at break change simulation on the state data of the hot-melt filamentary material based on the thermal viscosity decrease gradient data to obtain thermoplastic elongation at break change data; Step S233: analyzing the abnormal deformation contour parameters of the printed abnormal morphological deformation data to obtain the abnormal deformation contour parameters; Step S234: performing local abnormal deformation thermoplastic elongation distribution inversion on the abnormal deformation contour parameters according to the thermoplastic elongation at break change data to obtain the abnormal deformation thermoplastic distribution elongation; Step S235: performing abnormal calculation of printing retraction distance / speed based on abnormal deformation thermoplastic distribution elongation and abnormal deformation profile parameters to obtain abnormal printing retraction distance / speed; Step S236: reversely infer the abnormal wire drawing mechanism of the abnormal deformation profile parameters according to the abnormal printing retraction distance / speed and the abnormal deformation thermoplastic distribution elongation to obtain the abnormal wire drawing mechanism.

[0030] In the embodiment of the present invention, in the process of analyzing the thermal viscosity drop gradient of the state data of the hot-melt filamentary material, the thermal viscosity change curve is first constructed according to the material viscosity data under different temperature conditions. The viscosity of the material at different temperatures (180°C, 190°C, 200°C, 210°C, 220°C) is measured using a Brookfield rotational viscometer, and the viscosity change value with a shear rate range from 10s⁻¹ to 1000s⁻¹ is recorded. The relationship between thermal viscosity and temperature is fitted using the Arrhenius equation, and the thermal viscosity drop gradient of the material is calculated by differential operation. For example, in the PLA material, the viscosity at 200°C is measured to be 1850mPa·s, and it drops to 1120mPa·s at 210°C, and the thermal viscosity drop gradient is calculated to be about -73mPa·s / °C. According to the data changes in different temperature ranges, a thermal viscosity drop gradient data table is established and stored for subsequent calculations. In the process of simulating the change of thermoplastic elongation at break of hot-melt filamentary material state data based on the thermal viscosity drop gradient data, the viscosity data in the corresponding temperature range was first selected, and the change of elongation at break at different temperatures was calculated by combining the stress-strain experimental data of the material. The uniaxial tensile test of PLA material was carried out at 190℃, 200℃, 210℃ and 220℃ by Instron tensile testing machine, and the elongation at break of the material at different strain rates was recorded. For example, at 200℃, the elongation at break of PLA was measured to be 3.8%, which increased to 5.2% at 220℃. Based on the experimental data, the thermoplastic elongation at break change curve was constructed by cubic spline interpolation method, and the change rate of elongation at break was calculated by numerical differentiation. The calculated thermoplastic elongation at break change data was stored for subsequent abnormal deformation analysis. In the process of abnormal deformation contour parameter analysis of abnormal deformation data of printing, the abnormal morphology occurring during printing was measured by high-speed camera and image processing technology. First, the Canny edge detection algorithm was used to identify the edge of the printing trajectory and calculate the deviation amplitude of the contour. Secondly, the contour parameters of abnormal deformation are extracted by morphological analysis methods, including maximum deviation, local curvature change and deformation area. In the experiment, at a printing speed of 200mm / s, it was found that the maximum deviation of the contour of a certain printing path reached 0.35mm, which was significantly larger than the normal path (within 0.1mm). In addition, the local curvature distribution of the deformed area was calculated, and it was found that the curvature change in the abnormal area exceeded 0.05mm⁻¹. All abnormal deformation contour parameters are stored as structured data for subsequent analysis. In the process of inverting the thermoplastic elongation distribution of local abnormal deformation based on the thermoplastic elongation at break change data, the area affected by stretching is first extracted based on the spatial distribution of abnormal deformation. Then, combined with the thermoplastic elongation at break change data, the local stretch rate distribution of the abnormal area is calculated.The finite element method (FEM) was used to simulate the tensile deformation behavior of the abnormal area at different temperatures, the thermoplastic elongation at break data of the material was input, and the corresponding boundary conditions were applied for iterative calculation. For example, at 220°C, the local elongation of the abnormal area reached 4.5%, which was significantly higher than that of the normal area (2.8%). Finally, the distribution data of the thermoplastic elongation of abnormal deformation was obtained and stored for subsequent abnormal wire drawing analysis. In the process of abnormal calculation of the printing retraction distance / speed based on the thermoplastic distribution elongation of abnormal deformation and the abnormal deformation profile parameters, the critical conditions for the material to retract after stretching were first calculated. According to the viscoelastic characteristics of thermoplastic materials, a retraction mechanical model was established, and the Maxwell viscoelastic equation was used to calculate the retraction distance and speed of the material at different temperatures and stretching rates. For example, at 200°C, the retraction distance of PLA was measured to be 2.1mm, while at 220°C, the retraction distance increased to 3.6mm. Combined with the abnormal deformation profile parameters, the correlation between the retraction distance and the abnormal area was analyzed, and the retraction speed distribution was calculated. For example, in a specific abnormal area, the retraction speed reaches 0.8mm / s, which is higher than 0.5mm / s in the normal printing area. Finally, the abnormal printing retraction distance / speed data is generated and stored for abnormal wire drawing mechanism inversion. In the process of inverting the abnormal wire drawing mechanism of abnormal deformation profile parameters based on the abnormal printing retraction distance / speed and the abnormal deformation thermoplastic distribution elongation, the relationship between the retraction rate and the elongation rate in the wire drawing area is first analyzed, and the boundary conditions for wire drawing formation are calculated. The piecewise linear regression method is used to fit the retraction distance and elongation rate curve of the abnormal wire drawing area, and the wire drawing critical point is calculated. For example, in the printing process of PLA material, when the retraction speed exceeds 0.75mm / s and the local elongation rate exceeds 3.5%, the wire drawing phenomenon begins to appear. Combined with the printing abnormal morphological deformation data, the correlation between the wire drawing length and the nozzle motion trajectory is further analyzed, and the wire drawing length distribution and wire drawing diameter change are calculated. In the experiment, the wire drawing length was measured to be 4.2mm at 200℃, and increased to 6.8mm at 220℃. Based on all the calculated data, the abnormal wire drawing mechanism is derived and stored for subsequent warping probability calculation and printing failure detection.

[0031] Step S24 includes the following steps: Step S241: performing internal void distribution identification on the printed abnormal morphology deformation data to obtain abnormal morphology internal void distribution data; Step S242: reversely infer the change of hot melt tension of the material based on the abnormal wire drawing mechanism to obtain the change data of hot melt tension of the material; Step S243: Calculating the incremental gradient of the material hot melt cooling shrinkage rate according to the abnormal morphology internal void distribution data and the material hot melt tension change data to obtain the incremental gradient of the hot melt cooling shrinkage rate; Step S244: performing a multivariate linear regression analysis on the incremental gradient of the hot melt cooling shrinkage rate to obtain the incremental linear regression gradient of the shrinkage rate; Step S245: estimating the probability of warping between printing layers based on the linear regression gradient of the shrinkage increment to obtain the probability data of warping between printing layers.

[0032] In the embodiment of the present invention, in the process of identifying the internal void distribution of the abnormal morphological deformation data of the print, the internal structure data of the print is first obtained by industrial CT scanning technology, and the internal voids are identified and analyzed by image processing algorithm. The X-ray tube voltage of the industrial CT scanning equipment is set to 100kV, the exposure time is 500ms, and the resolution of the acquired tomographic image is set to 50μm. The density distribution of the print is layered using the Otsu threshold segmentation method, and the noise is removed by combining the morphological filtering method. For the identified low-density area, the connected region analysis algorithm is used to calculate the void area and distribution, and the void distribution data is stored. For example, in a PLA print, multiple internal voids are identified, the maximum void area is 2.3mm², the average void diameter is 0.8mm, and they are mainly distributed at the junction of the printing path. Combined with the void morphology data, the void volume and its distribution probability are further calculated for subsequent analysis. In the process of reversely inferring the change of hot melt tension of the material based on the abnormal wire drawing mechanism, the flow behavior of the material in the molten state during the printing process is first analyzed, and the relationship between the nozzle extrusion rate and the tensile force is calculated. Using a high-temperature tensile test device, uniaxial tensile tests were performed on PLA materials at three temperatures: 200°C, 210°C, and 220°C, and the tension changes at different tensile rates were recorded. Experimental data show that at 200°C and a tensile rate of 5mm / s, the material tension is 2.1N, and at 220°C and a tensile rate of 10mm / s, the tension rises to 3.8N. Based on these data, a nonlinear fitting method is used to calculate the rate of change of hot melt tension, and the tension is reversed in combination with the actual printing parameters of the abnormal wire drawing area. For example, in a certain abnormal wire drawing area, the extrusion speed is 50mm / s, the nozzle temperature is 210°C, and the corresponding hot melt tension is 3.2N calculated by interpolation, and the data is stored for subsequent calculations. In the process of calculating the incremental gradient of the hot melt cooling shrinkage rate of the material based on the internal void distribution data of the abnormal morphology and the material hot melt tension change data, the shrinkage behavior of the material during the cooling process is first calculated. The thermal expansion coefficient of the material was determined using a differential scanning calorimeter (DSC), and the expansion rate was recorded in the range of 100℃-200℃. The finite element analysis software was used to simulate thermal stress and calculate the cooling shrinkage change. Taking PLA material as an example, the shrinkage was measured to be 0.37% during the cooling process from 200℃ to room temperature, but in the abnormal void area, the shrinkage increased to 0.45% due to the decrease in local material density. Combined with the hot melt tension data, the shrinkage increment gradient was calculated. For example, when the internal void density reached 1.5 / cm³, the local shrinkage increment reached 0.08%. All calculation results were stored as a data table for subsequent analysis. In the process of multivariate linear regression analysis of the hot melt cooling shrinkage increment gradient, the shrinkage data under different printing conditions were first collected, including parameters such as nozzle temperature, printing speed, layer thickness, and internal void density, and a regression analysis matrix was constructed.The least squares method is used to regress and fit the relationship between the shrinkage increment and each parameter. For example, under the conditions of 200℃, printing speed of 60mm / s, layer thickness of 0.2mm, and internal void density of 1.5 / cm³, the regression analysis obtains the shrinkage increment prediction formula: ΔS=0.005T+0.002V+0.01D−0.0003H, where T is the nozzle temperature (℃), V is the printing speed (mm / s), D is the void density (pieces / cm³), and H is the layer thickness (mm). Finally, the linear regression gradient data of the shrinkage increment is obtained and stored for the calculation of the warping probability. In the process of estimating the probability of warping between printed layers based on the linear regression gradient of the shrinkage increment, the local stress distribution between different layers of the printed part is first calculated according to the regression gradient, and the thermal stress changes between printed layers are simulated using the linear elastic finite element method. The interlayer adhesion model is established using ABAQUS software, and the Young's modulus of the material is set to 3.5GPa, the Poisson's ratio is set to 0.36, and the temperature change condition is loaded to calculate the interlayer warping risk. For example, in the corner area of ​​a certain printed part, due to the local shrinkage increment of 0.09%, the calculated interlayer peeling stress is 2.1MPa, which exceeds the interlayer bonding strength of the PLA material (1.8MPa), so the probability of warping is as high as 85%. Finally, the interlayer warping probability data of the printed part is calculated and stored for subsequent printing failure detection.

[0033] Step S25 includes the following steps: Step S251: evaluating the non-uniform cooling rate of the inter-layer warping probability data to obtain the inter-layer warping non-uniform cooling rate; Step S252: performing numerical fluctuation simulation of internal stress accumulation on the probability data of interlayer warping according to the non-uniform cooling rate of interlayer warping to obtain numerical fluctuation data of interlayer internal stress; Step S253: performing interlayer adhesion instability risk interval analysis based on interlayer internal stress numerical fluctuation data and interlayer warping non-uniform cooling rate to obtain an interlayer adhesion instability risk interval; Step S254: performing cracking risk calculation according to the interlayer adhesion instability risk interval to obtain interlayer cracking risk data.

[0034] In an embodiment of the present invention, in the process of evaluating the non-uniform cooling rate of the probability data of warping between printed layers, the temperature change of the printed layer is first monitored in real time based on a thermal imager, and the cooling rate at different positions is calculated. Using a FLIRA655sc thermal imager, the temperature field data of the printed layer is recorded at a sampling frequency of 10 Hz during the printing process, and the temperature gradient changes of the printed layer at different time points are analyzed. The cooling rate is calculated using the second-order finite difference method, and a spatial coordinate system is established to map the cooling rate to different areas between printed layers. For example, in the corner area of ​​a certain print, the cooling rate is 0.12°C / s, while the cooling rate in the middle area is only 0.05°C / s. Through the non-uniform cooling rate evaluation, the temperature gradient distribution between layers is calculated, and the data is stored for subsequent analysis. In the process of simulating the numerical fluctuation of internal stress accumulation of the probability data of warping between printed layers according to the non-uniform cooling rate of the warping between layers, the thermal stress-structure coupling simulation method is first used to calculate the internal stress changes caused by thermal contraction during the interlayer cooling process. Using ABAQUS finite element analysis software, a temperature field-stress field coupling model between printed layers is constructed, and the thermal expansion coefficient of the PLA material is set to , Young's modulus is 3.5GPa, and Poisson's ratio is 0.36. In the cooling stage, temperature change boundary conditions are applied to calculate the stress accumulation rate in different areas. For example, in the edge area of ​​the printed part, due to the high cooling rate, the calculated interlayer internal stress peak reaches 4.2MPa, while in the central area, it is only 2.3MPa. The time domain Fourier transform (FFT) method is further used to analyze the stress numerical fluctuation characteristics, extract the frequency components of the interlayer stress change, and store the results for subsequent risk analysis. In the process of analyzing the interlayer adhesion instability risk interval based on the interlayer internal stress numerical fluctuation data and the interlayer warping non-uniform cooling rate, the extreme value distribution of the interlayer shear stress is first calculated, and the safety interval of the interlayer adhesion is analyzed based on the critical instability theory. The interlayer bonding strength is calculated using the friction slip model, the interlayer shear strength threshold of the PLA material is set to 2.1MPa, and the extreme value statistical method is used to calculate the interlayer shear stress peak distribution under different cooling rates. For example, in a certain printing area, the maximum shear stress is 2.3MPa, which exceeds the material adhesion threshold and is therefore marked as a high-risk area. At the same time, the probability of interlayer adhesion instability under different printing conditions is calculated by Monte Carlo random simulation method. For example, when the cooling rate changes from 0.08℃ / s to 0.15℃ / s, the probability of instability reaches 72%. Finally, the interlayer adhesion instability risk interval data is output and stored for cracking risk calculation. In the process of cracking risk calculation according to the interlayer adhesion instability risk interval, the critical stress intensity factor of interlayer cracking is first calculated based on fracture mechanics theory, and the crack propagation risk is evaluated in combination with the stress fluctuation characteristics of the interlayer. The critical stress intensity factor of PLA material is calculated by Griffith fracture criterion, the crack initiation length is set to 100μm, and the critical stress intensity threshold is calculated to be 0.72MPa·m¹ / ². The extended finite element method (XFEM) is further used to simulate the interlayer crack propagation process and calculate the crack growth rate. For example, in the high warping area, the simulated crack propagation rate is 0.005mm / s, while in the low warping area, the crack propagation rate is only 0.001mm / s. Combining the crack propagation rate and interlayer stress fluctuation characteristics, the overall cracking risk between printed layers is calculated, and finally the interlayer cracking risk data is generated and stored for use in subsequent printing failure detection systems.

[0035] Step S253 includes the following steps: The growth trend of the interlaminar internal stress numerical fluctuation data is numerically quantified to obtain the growth trend numerical quantification data; According to the growth trend numerical quantification data, the monotonically increasing mutation point distribution of the interlaminar internal stress numerical fluctuation data is identified to obtain the internal stress increasing mutation point distribution data; According to the distribution data of the internal stress increase mutation point and the non-uniform cooling rate of the interlayer warping, the non-linear relationship coupling data of the internal stress mutation-cooling rate relationship is obtained; The segmented local trend convergence of the coupled data of the internal stress mutation-cooling rate relationship is performed to obtain the local convergence trend of the internal stress mutation-cooling rate; Based on the Newton iteration method, the local convergence trend of internal stress mutation-cooling rate is numerically iterated locally to obtain the convergence iteration value of internal stress mutation-cooling rate; The interlayer adhesion instability risk interval analysis was performed based on the internal stress mutation-cooling rate convergence iteration value, and the interlayer adhesion instability risk interval was obtained.

[0036] In the embodiment of the present invention, in the process of numerically quantifying the growth trend of the interlaminar internal stress numerical fluctuation data, firstly, based on the time series stress data, the growth rate in different time windows is calculated, and the numerical expression of the growth trend is obtained by using the numerical differentiation method. The stress fluctuation data of each time step is calculated by the central difference method, and the time step is set to 0.5s to calculate the internal stress growth rate of each time interval. For example, the stress data in a certain area is [2.1MPa, 2.3MPa, 2.7MPa, 3.2MPa], and the growth trend data [0.2MPa / s, 0.4MPa / s, 0.5MPa / s] are obtained by numerical differentiation. The growth trend data is further smoothed by using a sliding average filter to eliminate local abnormal disturbances, and finally stable growth trend numerical quantification data is obtained. In the process of identifying the distribution of monotonically increasing mutation points of the interlaminar internal stress numerical fluctuation data according to the growth trend numerical quantification data, the second-order derivative of the stress growth curve is first calculated to determine the position of the mutation point. Lagrange interpolation is used to fit the growth trend data and calculate the curvature change rate at each time point. When the second-order derivative exceeds the set threshold (for example, 0.1MPa / s²), it is marked as a mutation point. For example, in the stress growth curve of a certain area, mutation points are detected at 2s and 4.5s, indicating that the area has undergone significant stress accumulation changes at this time point. Finally, the positions of all mutation points and the corresponding stress values ​​are stored to form the distribution data of internal stress increasing mutation points. In the process of coupling the nonlinear relationship between the distribution data of internal stress increasing mutation points and the non-uniform cooling rate of interlayer warping, the Pearson correlation coefficient is used to calculate the linear correlation between the two, and support vector regression (SVR) is used for nonlinear regression fitting. The kernel function is set to radial basis function (RBF), the regularization parameter C=1.0, and the relationship curve between internal stress mutation points and cooling rate is fitted. For example, in the corner area of ​​a certain print, the stress mutation point distribution data is [3.2MPa, 3.5MPa, 3.9MPa], and the corresponding cooling rate data is [0.10℃ / s, 0.12℃ / s, 0.15℃ / s]. The nonlinear coupling equation is obtained by SVR fitting, and finally the internal stress mutation-cooling rate relationship coupling data is formed. In the process of segmented local trend convergence of the internal stress mutation-cooling rate relationship coupling data, the double sample t test is first used to detect whether there is a significant difference in the stress growth trend in different cooling rate intervals. The entire data set is divided into several sub-intervals according to the cooling rate, such as [0.08℃ / s, 0.10℃ / s], [0.10℃ / s, 0.12℃ / s], etc. The mean of the stress growth rate is calculated in each interval, and a hypothesis test is performed. If the P value is less than 0.05, it is determined that there is a significant change in the trend in the interval. Finally, trend convergence analysis is performed on all intervals, and local convergence trend data is stored.In the process of local convergence numerical iteration of the local convergence trend of internal stress mutation-cooling rate based on Newton iteration method, a nonlinear equation of cooling rate and stress growth is first constructed, and the roots of the equation are solved using Newton iteration method. The initial value is set to the average value of the cooling rate (for example, 0.11℃ / s), and the update value of each step is calculated, and the error limit is used. As the convergence judgment condition. For example, in one iterative calculation, the initial value is set to 0.11, the calculated next value is 0.108, and the iteration continues to 0.107. After the error meets the convergence condition, the internal stress mutation-cooling rate convergence iteration value is finally obtained. In the process of analyzing the interlayer adhesion instability risk interval according to the internal stress mutation-cooling rate convergence iteration value, the interlayer adhesion strength threshold is first calculated, and the critical point of instability is predicted based on the convergence iteration value. The interlayer shear strength of the PLA material is set to 2.1MPa, and the Bayesian probability analysis method is used to calculate the instability risk probability at different cooling rates. For example, when the cooling rate is 0.12℃ / s, the interlayer instability probability is calculated to be 65%. When the cooling rate is reduced to 0.09℃ / s, the instability probability is reduced to 28%. Finally, the interlayer adhesion instability risk interval data is stored and used for subsequent cracking risk analysis. Among them, it is necessary to explain that "internal stress mutation-cooling rate convergence iteration value" refers to the coupling relationship between interlayer internal stress mutation and cooling rate quantified by numerical calculation method during 3D printing, and its final stable value is solved by iterative method. During the cooling process of the printed layer, the cooling rates in different regions are different, resulting in uneven accumulation of internal stress between layers, and mutations in some areas. There is a complex nonlinear relationship between this mutation point and the cooling rate. In order to accurately analyze this relationship, the growth trend of interlayer internal stress fluctuations is first identified by numerical methods, and the position and intensity of mutation points are detected. Then, the mathematical model of stress mutation points and cooling rate is established by nonlinear regression method, and local trend analysis is performed to determine its convergence characteristics. On this basis, the local convergence trend of internal stress mutation-cooling rate is numerically optimized by Newton iteration method, that is, starting from the initial estimated value, it is gradually approached to the final stable coupling value through continuous iterative updates, ensuring that the calculation results can reflect the actual impact of cooling rate on internal stress mutation. The final "converged iterative value" is a stable mathematical solution, which represents the numerical characteristics of the sudden change of interlaminar internal stress under different cooling rate conditions and can be used to further predict the instability risk of interlaminar adhesion.

[0037] In another embodiment, when the growth trend of the interlaminar internal stress numerical fluctuation data is numerically quantified, the data must first be analyzed in time series to identify the trend of the internal stress fluctuation in the data. By performing sliding average processing on the time series data, noise interference is removed and the basic trend of the fluctuation is retained. The growth change between the data points is quantified using the slope calculation method to determine the growth rate at each time point. In this way, the obvious growth trend in the internal stress fluctuation process can be clearly extracted. A threshold is set, and when the growth exceeds the threshold, it is marked as important growth trend data. The data quantified by the growth trend will serve as the basis for subsequent analysis, providing the growth rate of each data point and its change law in the time series, providing data support for subsequent mutation point detection and nonlinear coupling. According to the growth trend numerical quantification data, when the distribution of monotonically increasing mutation points is identified for the interlaminar internal stress numerical fluctuation data, the quantified growth trend data is first analyzed to find the monotonically increasing part in the data. The location where the growth trend in the data mutates is identified by differential calculation or a mutation point detection algorithm based on local extreme values. The identification of mutation points is based on the fact that the growth trend rises or falls rapidly within a certain interval, and the mutation point is marked when the rate of change exceeds the set threshold. The sliding window algorithm ensures that the internal stress mutation that changes with time can be tracked dynamically. In the process of identifying the distribution of mutation points, the long-term and short-term dependencies of the time series are considered to avoid misidentification or missed identification. The distribution data of the internal stress increase mutation points finally obtained provide the location and intensity of the mutation in the process of internal stress fluctuation, and further provide key data support for the nonlinear coupling analysis of internal stress and cooling rate. When the nonlinear relationship coupling is performed based on the distribution data of the internal stress increase mutation points and the non-uniform cooling rate of the interlayer warping, it is first necessary to associate the location of the internal stress mutation point with the change of the cooling rate. The cooling rate data corresponding to different time points are obtained by interpolation method, especially the rate of change of the cooling rate. Then the polynomial regression method is used to establish the nonlinear relationship between the internal stress mutation and the cooling rate, and the appropriate order is selected for fitting to capture the complex mutual influence between the two. In order to enhance the accuracy of the model, the regularization method is introduced to avoid overfitting. In this process, the occurrence of internal stress mutation and the change of cooling rate have a significant mutual influence. The coupling analysis can effectively reveal the sensitivity of internal stress mutation under different cooling conditions and provide a detailed data model for subsequent risk analysis. When performing segmented local trend convergence on the coupled data of the internal stress mutation-cooling rate relationship, the coupled data is first divided into several sub-intervals according to the cooling rate change interval, and the data in each sub-interval has similar cooling rate characteristics. Then, in each sub-interval, the local convergence trend of the data is calculated, that is, the data change trend of each sub-interval is fitted by the local regression analysis method to evaluate the stability of its change.Local trend convergence analysis helps to determine whether there is a certain regular trend convergence in the occurrence of internal stress mutation under specific cooling rate conditions. During the analysis process, the window function is used to optimize the convergence process and ensure that the trend convergence degree of each sub-interval is high, so as to provide stable initial conditions for subsequent numerical iterations. When the local convergence numerical iteration of the local convergence trend of internal stress mutation-cooling rate is performed based on the Newton iteration method, the local trend is refined and optimized by using the Newton iteration method. The initial convergence value is set and its deviation is calculated, and the accuracy is improved by gradually approaching the true value. In each iteration, based on the value of the current convergence trend, the change rate of the local trend is calculated by derivation, and the convergence value is continuously adjusted until the set accuracy standard is met. During the iteration process, it is necessary to select an appropriate step size to ensure the stability of the calculation process and avoid numerical divergence or non-convergence of calculation due to excessive step size. Through multiple iterations, a stable internal stress mutation-cooling rate convergence iteration value is finally obtained, providing accurate data support for subsequent risk interval analysis. When analyzing the risk interval of interlayer adhesion instability based on the internal stress mutation-cooling rate convergence iteration value, the risk analysis model is first constructed based on the converged value. The adhesion change corresponding to the convergence value of the internal stress mutation and the cooling rate is modeled using the statistical analysis method. The stability of the adhesion under different internal stress levels is analyzed through the nonlinear relationship between the interlayer adhesion and the internal stress mutation-cooling rate. When the coupling value of the internal stress mutation and the cooling rate exceeds the set instability threshold, the risk interval of interlayer adhesion instability is predicted. At this time, the critical value of interlayer adhesion instability under different internal stress and cooling rate conditions is clarified through the interval analysis method. The final interlayer adhesion instability risk interval provides an early warning for failures in the 3D printing process, which helps to monitor and adjust parameters in real time during the printing process, thereby improving the printing quality and success rate.

[0038] Step S3 includes the following steps: Step S31: performing convolution calculation on the risk level of cracking between printed layers to obtain convolution data of the risk of cracking between printed layers; Step S32: performing logic learning on the risk convolution data of cracking between printed layers to obtain logic data of cracking risk between printed layers; Step S33: constructing a printing layer crack risk identification model based on the printing layer crack risk logic data based on the convolutional neural network to obtain the printing layer crack risk identification model.

[0039] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes: Step S31: performing convolution calculation on the risk level of cracking between printed layers to obtain convolution data of the risk of cracking between printed layers; In the embodiment of the present invention, during the convolution calculation of the risk level of cracking between printed layers, the risk data of cracking between printed layers is first obtained, and the data is discretized to construct a multi-dimensional tensor data set. Based on the material characteristics of 3D printing, the thickness distribution, cooling rate, thermal expansion coefficient and stress accumulation data between printed layers are set as input variables, and the data is normalized so that the data values ​​are distributed in the range of 0 to 1 to improve the calculation accuracy. The interlayer data is convoluted using a three-dimensional convolution kernel, the convolution kernel size is set to 3×3×3, and a sliding window with a step size of 1 is used to scan the entire data set layer by layer. The stress gradient change characteristics between different layers are extracted through convolution calculation, and the negative value data is truncated using the ReLU activation function to ensure the effectiveness of the nonlinear characteristics during the calculation process. Finally, the convolution data of the risk of cracking between printed layers is generated, which contains the risk distribution of different areas between printed layers and is used for subsequent logical analysis.

[0040] Step S32: performing logic learning on the risk convolution data of cracking between printed layers to obtain logic data of cracking risk between printed layers; In an embodiment of the present invention, in the process of logical learning of the convolution data of the risk of cracking between printed layers, the convolution data is first converted into a matrix representation, and the data is classified and learned using a decision tree method. The cracking risk categories between different layers are set, including low risk, medium risk and high risk categories, and a decision tree model is constructed based on the existing cracking history data. The information gain algorithm is used to calculate the weights of different features, and the feature with the highest information gain is selected as the decision node. For example, in a batch of printing data, the correlation coefficient between the interlayer cooling rate and the cracking probability reaches 0.85, so this feature is preferentially selected into the decision path. The C4.5 algorithm is further used to prune the decision tree, remove redundant nodes, and improve the generalization ability of logical learning. Finally, the logical data of the risk of cracking between printed layers is formed, which is used to construct a cracking risk identification model.

[0041] Step S33: constructing a printing layer crack risk identification model based on the printing layer crack risk logic data based on the convolutional neural network to obtain the printing layer crack risk identification model.

[0042] In the embodiment of the present invention, in the process of constructing a model for identifying the risk of cracking between printed layers based on the logical data of the risk of cracking between printed layers using a convolutional neural network, a neural network structure including multiple convolutional layers and pooling layers is first constructed, and the dimension of the input data is set to 100×100×10 to adapt to the distribution of different printed layers. The convolution kernel size of the first convolutional layer is set to 5×5, the number of channels is set to 16, and the maximum pooling layer is used to reduce the feature dimension to reduce the computational complexity. Subsequently, the second convolutional layer is introduced, the convolution kernel size is set to 3×3, the number of channels is increased to 32, and the batch normalization method is used to accelerate the convergence process. In the fully connected layer stage, the Softmax activation function is used to calculate the risk probability, and the model is optimized based on the cross entropy loss function. The Adam optimizer is used to adjust the learning rate, and the initial learning rate is set to 0.001, which decays to 0.9 times the original after every 10 rounds of training. Finally, a model for identifying the risk of cracking between printed layers was obtained. This model can predict the cracking risk between different layers based on the input printing parameters and interlayer stress data, and provide data support for quality control in the 3D printing process.

[0043] Step S33 includes the following steps: Step S331: extracting multi-scale features from the printed interlayer crack risk logic data to obtain interlayer crack multi-scale risk data; Step S332: dividing the interlayer cracking multi-scale risk data into a training set and a test set, and obtaining an interlayer cracking multi-scale risk training set and an interlayer cracking multi-scale risk test set respectively; Step S333: constructing an initial printing interlayer crack risk identification model for the interlayer crack multi-scale risk training set based on a convolutional neural network to obtain an initial printing interlayer crack risk identification model; Step S334: performing a model test on the initial printing interlayer crack risk identification model according to the interlayer crack multi-scale risk test set to obtain a printing interlayer crack risk identification model.

[0044] In the embodiment of the present invention, in the process of multi-scale feature extraction of the risk logic data of interlayer cracking, the data is firstly analyzed based on convolution kernels of different scales to extract the risk features of cracking at different levels between the printed layers. The convolution kernel sizes are set to 3×3, 5×5 and 7×7 respectively, and the data is convoluted by sliding windows to extract local, regional and global interlayer risk features. For 3D printing materials with different layer thicknesses, a dynamic pooling strategy is adopted to enable feature extraction to adapt to different printing conditions. For thinner interlayer areas, a smaller pooling window, such as a 2×2 pooling window, is used to retain more detailed information; for thicker interlayer areas, a 4×4 pooling window is used to reduce the data dimension and improve the calculation efficiency. In this way, the multi-scale risk data of interlayer cracking is obtained, which contains the risk change characteristics of different scales during the printing process and provides data support for subsequent training and testing. In the process of dividing the multi-scale risk data of interlayer cracking into training sets and test sets, the data set is firstly randomized, and a stratified sampling method is used to ensure that data of different risk levels are evenly distributed. In the division process, 80% of the data is set as the training set and 20% of the data is set as the test set, and the statistical characteristics of the training set and the test set in terms of interlayer thickness, cooling rate, stress distribution, etc. are kept consistent. The K-fold cross-validation method is used to further divide the validation set within the training set to ensure that the model has strong generalization ability during the training process. For abnormal data points, such as interlayer warping exceeding the set threshold (such as 0.5mm) or local stress concentration exceeding the material yield limit (such as 50MPa), data enhancement methods are used to expand them to enhance the model's ability to recognize extreme situations. Finally, the multi-scale risk training set of interlayer cracking and the multi-scale risk test set of interlayer cracking are obtained to provide data input for subsequent model training and testing. In the process of constructing the interlayer cracking risk identification model based on the initial printing of the multi-scale risk training set of interlayer cracking based on the convolutional neural network, a deep convolutional neural network is first constructed, including multiple convolutional layers, pooling layers, and fully connected layers. The input layer receives the normalized training data. The first convolution layer uses 16 3×3 convolution kernels, and the second convolution layer uses 32 5×5 convolution kernels to extract local and medium-scale risk features between layers. Then the Batch Normalization layer is added to normalize the data of different batches to accelerate the training convergence speed. The pooling layer adopts the maximum pooling strategy to reduce the data dimension and improve the computational efficiency. In the fully connected layer stage, the Dropout method is introduced, and the dropout rate is set to 0.5 to prevent the model from overfitting. Finally, the Softmax activation function is used to classify different risk levels, and the cross entropy loss function is used to calculate the error. The Adam optimization algorithm is used to adjust the learning rate. The initial learning rate is set to 0.001 and dynamically adjusted during the training process.After the training is completed, the initial inter-layer cracking risk identification model is obtained, which can preliminarily identify the cracking risk level between different printed layers. In the process of model testing the initial inter-layer cracking risk identification model according to the inter-layer cracking multi-scale risk test set, the test set data is first used as input, and the prediction results output by the model are compared and analyzed with the actual annotated risk level. The classification accuracy of the model is calculated using the confusion matrix, and the model performance is evaluated using indicators such as Precision, Recall, and F1-score. For the misclassified samples, error analysis is performed, and the feature extraction process of the convolutional neural network is visualized using the Grad-CAM method to observe the focus of the model on different inter-layer features. During the test process, if it is found that the model has a high misclassification rate under a specific layer thickness or cooling rate, the data rebalancing method is used to increase the amount of data in this category and perform incremental training. After multiple rounds of testing and optimization, the inter-layer cracking risk identification model is finally obtained, which can accurately predict the cracking risk between printed layers and provide reliable data support for 3D printing failure detection.

[0045] The present invention also provides a 3D printer printing failure detection system, which is used to execute the 3D printer printing failure detection method as described above, and the 3D printer printing failure detection system comprises: The abnormal state extraction module is used to obtain the operation log of the 3D printer; perform fused deposition modeling abnormal state extraction on the operation log of the 3D printer to obtain fused deposition modeling abnormal state data; The cracking risk calculation module is used to reverse the abnormal wire drawing mechanism based on the abnormal state data of fused deposition modeling to obtain the abnormal wire drawing mechanism; estimate the probability of warping between printing layers based on the abnormal wire drawing mechanism to obtain the probability data of warping between printing layers; perform cracking risk calculation on the probability data of warping between printing layers to obtain the risk data of cracking between printing layers; A cracking risk identification model building module is used to build a printing interlayer cracking risk identification model based on the printing interlayer cracking risk data based on a convolutional neural network to obtain a printing interlayer cracking risk identification model; The detection execution module is used to send the printing layer crack risk identification model to the 3D printer control center to perform printing failure detection.

[0046] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A 3D printer printing failure detection method, characterized in that: The following steps are involved: Step S1: Obtaining the 3D printer operation log; Extract abnormal state of fused deposition modeling from the operation log of the 3D printer to obtain abnormal state data of fused deposition modeling; Wherein, step S2 comprises: Step S21: obtaining hot-melt filament material state data; Step S22: performing printing abnormal morphological deformation identification based on the fused deposition modeling abnormal state data to obtain printing abnormal morphological deformation data; Step S23: reversely inferring the abnormal wire drawing mechanism of the printed abnormal morphological deformation data according to the state data of the hot-melt filamentary material to obtain the abnormal wire drawing mechanism; Step S24: estimating the probability of warping between printing layers based on the printing abnormal morphological deformation data and the abnormal wire drawing mechanism, and obtaining the probability data of warping between printing layers; Step S25: performing cracking risk calculation on the probability data of warping between printed layers to obtain cracking risk data between printed layers; Step S3: constructing a printing layer crack risk identification model based on the printing layer crack risk data based on the convolutional neural network to obtain a printing layer crack risk identification model; Step S4: Sending the inter-layer crack risk identification model to the 3D printer control center to perform printing failure detection.

2. The 3D printer printing failure detection method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Obtaining the 3D printer operation log; Step S12: performing data cleaning on the 3D printer operation log to obtain a 3D printer operation cleaning log; Step S13: extracting abnormal state of fused deposition modeling from the cleaning log of the 3D printer to obtain abnormal state data of fused deposition modeling.

3. The 3D printer printing failure detection method according to claim 1, characterized in that: Step S23 includes the following steps: Step S231: performing a thermal viscosity drop gradient analysis on the state data of the hot-melt filamentary material to obtain thermal viscosity drop gradient data; Step S232: performing a thermoplastic elongation at break change simulation on the state data of the hot-melt filamentary material based on the thermal viscosity decrease gradient data to obtain thermoplastic elongation at break change data; Step S233: analyzing the abnormal deformation contour parameters of the printed abnormal morphological deformation data to obtain the abnormal deformation contour parameters; Step S234: performing local abnormal deformation thermoplastic elongation distribution inversion on the abnormal deformation contour parameters according to the thermoplastic elongation at break change data to obtain the abnormal deformation thermoplastic distribution elongation; Step S235: performing abnormal calculation of printing retraction distance / speed based on abnormal deformation thermoplastic distribution elongation and abnormal deformation profile parameters to obtain abnormal printing retraction distance / speed; Step S236: reversely infer the abnormal wire drawing mechanism of the abnormal deformation profile parameters according to the abnormal printing retraction distance / speed and the abnormal deformation thermoplastic distribution elongation to obtain the abnormal wire drawing mechanism.

4. The 3D printer printing failure detection method according to claim 1, characterized in that: Step S24 includes the following steps: Step S241: performing internal void distribution identification on the printed abnormal morphology deformation data to obtain abnormal morphology internal void distribution data; Step S242: reversely infer the change of hot melt tension of the material based on the abnormal wire drawing mechanism to obtain the change data of hot melt tension of the material; Step S243: Calculate the incremental gradient of the material hot melt cooling shrinkage rate according to the abnormal morphology internal void distribution data and the material hot melt tension change data to obtain the incremental gradient of the hot melt cooling shrinkage rate; Step S244: performing a multivariate linear regression analysis on the incremental gradient of the hot melt cooling shrinkage rate to obtain the incremental linear regression gradient of the shrinkage rate; Step S245: estimating the probability of warping between printing layers based on the linear regression gradient of the shrinkage increment to obtain the probability data of warping between printing layers.

5. The 3D printer printing failure detection method according to claim 1, characterized in that: Step S25 includes the following steps: Step S251: evaluating the non-uniform cooling rate of the inter-layer warping probability data to obtain the inter-layer warping non-uniform cooling rate; Step S252: performing numerical fluctuation simulation of internal stress accumulation on the probability data of interlayer warping according to the non-uniform cooling rate of interlayer warping to obtain numerical fluctuation data of interlayer internal stress; Step S253: performing interlayer adhesion instability risk interval analysis based on interlayer internal stress numerical fluctuation data and interlayer warping non-uniform cooling rate to obtain an interlayer adhesion instability risk interval; Step S254: performing cracking risk calculation according to the interlayer adhesion instability risk interval to obtain interlayer cracking risk data.

6. The 3D printer printing failure detection method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing convolution calculation on the risk level of cracking between printed layers to obtain convolution data of the risk of cracking between printed layers; Step S32: performing logic learning on the risk convolution data of cracking between printed layers to obtain logic data of cracking risk between printed layers; Step S33: constructing a printing layer crack risk identification model based on the printing layer crack risk logic data based on the convolutional neural network to obtain the printing layer crack risk identification model.

7. The 3D printer printing failure detection method according to claim 6, characterized in that: Step S33 includes the following steps: Step S331: extracting multi-scale features from the printed interlayer crack risk logic data to obtain interlayer crack multi-scale risk data; Step S332: dividing the interlayer cracking multi-scale risk data into a training set and a test set, and obtaining an interlayer cracking multi-scale risk training set and an interlayer cracking multi-scale risk test set respectively; Step S333: constructing an initial printing interlayer crack risk identification model for the interlayer crack multi-scale risk training set based on a convolutional neural network to obtain an initial printing interlayer crack risk identification model; Step S334: performing a model test on the initial printing interlayer crack risk identification model according to the interlayer crack multi-scale risk test set to obtain a printing interlayer crack risk identification model.

8. A 3D printer printing failure detection system, characterized in that: Used to execute the 3D printer printing failure detection method according to claim 1, the 3D printer printing failure detection system comprises: The abnormal state extraction module is used to obtain the operation log of the 3D printer; perform fused deposition modeling abnormal state extraction on the operation log of the 3D printer to obtain fused deposition modeling abnormal state data; The cracking risk calculation module is used to reverse the abnormal wire drawing mechanism based on the abnormal state data of fused deposition modeling to obtain the abnormal wire drawing mechanism; estimate the probability of warping between printing layers based on the abnormal wire drawing mechanism to obtain the probability data of warping between printing layers; perform cracking risk calculation on the probability data of warping between printing layers to obtain the risk data of cracking between printing layers; A cracking risk identification model building module is used to build a printing interlayer cracking risk identification model based on the printing interlayer cracking risk data based on a convolutional neural network to obtain a printing interlayer cracking risk identification model; The detection execution module is used to send the risk identification model of cracking between printing layers to the 3D printer control center to perform printing failure detection.

Images