Additive temperature monitoring and defect prediction method and system based on colorimetric temperature measurement principle

Abstract

The invention provides an additive material temperature monitoring and defect prediction method and system based on a colorimetric temperature measurement principle. The method comprises the steps that a position signal and a temperature signal of a motion system in the printing process are recorded; the method comprises the following steps: dividing a temperature signal collected according to a colorimetric temperature measurement principle into signal sections with the same time interval and consistent time length, and extracting characteristic parameters of temperature data in each section; scanning the printed additive manufacturing experiment sample to obtain internal defect information of the sample, and labeling each section of temperature signal data in combination with position signal data of a motion system; a deep neural network model is established, and position signals and temperature signals, obtained in real time in the machining process, of the motion system are transmitted into the trained DNN model; and the DNN model is used for automatically monitoring the temperature of a molten pool in the material adding process and predicting defects. The method improves the training precision, and obtains a model with a better prediction effect.

Description

Technical Field

[0001] The present invention relates to the field of additive manufacturing and artificial intelligence interaction technology, and specifically, to a method and system for additive temperature monitoring and defect prediction based on the colorimetric temperature measurement principle; more specifically, to a method and system for additive process molten pool temperature monitoring and defect prediction based on the colorimetric temperature measurement principle. Background Art

[0002] In the additive manufacturing process, the use of heterogeneous materials and gradient materials often leads to defects, which will affect the quality and performance of the manufactured parts. In the actual additive manufacturing process, the prediction and treatment of these internal and external defects become very important. By obtaining the temperature signal in the additive manufacturing process and combining it with appropriate algorithms, the identification and prediction of defects can be achieved.

[0003] In the process of additive manufacturing, the melting, solidification and curing of materials will produce complex temperature changes. These temperature signals contain rich information and can be used to reflect the melting state, solidification process and possible defects of the material. By collecting, analyzing and processing these temperature signals, defects can be identified and predicted, so that processing parameters can be adjusted in time or measures can be taken to avoid defects.

[0004] Combining advanced sensing technology and machine learning algorithms, the temperature signals obtained during additive manufacturing can be used for defect identification and prediction. This approach can not only improve the manufacturing quality and efficiency of additive manufacturing, but also better understand and control the defect formation mechanism during additive manufacturing. Effectively using temperature signals for defect identification and prediction has important theoretical and practical significance in the field of additive manufacturing. Summary of the invention

[0005] In view of the defects in the prior art, the object of the present invention is to provide an additive temperature monitoring and defect prediction method and system based on the colorimetric temperature measurement principle.

[0006] According to the present invention, a method for temperature monitoring and defect prediction of additive materials based on the colorimetric temperature measurement principle is provided, comprising:

[0007] Step S1: performing additive manufacturing experiments using different process parameters, and recording the position signal and temperature signal of the motion system during the printing process;

[0008] Step S2: dividing the temperature signal collected according to the colorimetric temperature measurement principle into signal segments with the same time interval and the same length, each signal segment contains the same amount of temperature data, and extracting characteristic parameters of the temperature data in each segment;

[0009] Step S3: Scan the printed additive manufacturing experimental sample to obtain the internal defect information of the sample, and label each section of the temperature signal data in combination with the position signal data of the motion system;

[0010] Step S4: Establish a deep neural network model, divide the data set into a training set and a test set for training, and save the trained DNN model and parameters; perform laser additive manufacturing, and transfer the position signal and temperature signal of the motion system obtained in real time during the manufacturing process to the trained DNN model;

[0011] Step S5: Use the DNN model to automatically monitor the temperature of the molten pool during the additive process and predict defects.

[0012] Preferably, in step S1:

[0013] Multiple additive manufacturing experiments were conducted using preset optimal process parameters and manually adjusted secondary process parameters; the process parameters included laser power, scanning speed, and powder feeding amount;

[0014] The details of the additive manufacturing experiment are as follows:

[0015] Explore the optimal process parameters. The optimal process parameters can print components that meet the preset standards. When different process parameters deviate from the values ​​of the optimal process parameters by too much or too little, defects will appear inside the printed components. Design experiments including optimal process parameter combinations and secondary process parameter combinations, covering a variety of situations, build data sets, print single-channel, thin-walled, block or other components according to the designed single-factor experiments, and collect position signals and temperature signals of the motion system during the printing process;

[0016] The temperature signal collected according to the principle of colorimetric temperature measurement is first converted into an analog signal and transmitted to the acquisition card. The acquisition card transmits the analog signal to the PC end, and LabView software is used in the PC end to convert the analog signal into temperature data.

[0017] Preferably, in step S2:

[0018] The complete temperature data is divided into signal segments with the same time interval and the same length. Adjacent temperature segments contain some of the same temperature data. The characteristic parameters of the temperature data in each segment are extracted, including the highest temperature T in the segment. max , minimum temperature T min , maximum temperature difference ΔT, maximum temperature change rate δT;

[0019] The temperature data characteristic parameters are calculated by the following formula:

[0020] ΔT=T max -T min

[0021]

[0022] Among them, T max is the maximum temperature, T min is the minimum temperature.

[0023] Preferably, in step S3:

[0024] The printed additive manufacturing experimental samples were subjected to electronic computer tomography to obtain the internal defect information of the samples, and each segment of temperature signal data was labeled in combination with the obtained motion system position signal data;

[0025] Internal defect information includes three categories: pores, cracks, and intact. Labels are manually marked after the additive manufacturing experiment is completed. By obtaining the position signal data and temperature signal data of the motion system during the printing process, the defective parts in the formed parts are matched with the temperature signal data one by one, and the defect conditions of the areas corresponding to different temperature signal sections are manually judged, and one of the three labels of pores, cracks, and intact is marked.

[0026] The labels include a normal defect-free state label, a crack defect state label and a pore defect state label.

[0027] Preferably, in step S4:

[0028] Establish a deep neural network DNN model, divide the data set into training set and test set, then train it, and save the trained DNN model and parameters;

[0029] The data set includes the input and output of the model. The input is the temperature signal paragraph data, and the output is the label corresponding to the temperature signal paragraph.

[0030] The neural network layers in the DNN model are divided into input layer, hidden layer and output layer. The layers are fully connected. The input layer inputs the highest temperature T max , minimum temperature T min , maximum temperature difference ΔT and maximum temperature change rate δT; the output layer outputs the normal defect-free state label, crack defect state label and pore defect state label; the hidden layer contains multiple layers of neurons, and the number of neurons and layers are adjusted according to the actual data set;

[0031] The input between the fully connected layers passes through the activation function, and the activation function before the output layer is set to SoftMax, and its calculation formula is:

[0032]

[0033] Among them, t represents an array; t i Represents the i-th element of the array; e is a natural constant;

[0034] The activation function of the fully connected layer before the output layer is set to Sigmoid or ReLU function, which is selected and adjusted according to the situation. The calculation formulas are:

[0035]

[0036] ReLU(t)=max(0,t)

[0037] The choice of activation function is adjusted according to the actual operation results and specific loss conditions. When the output result deviates from 0, the Sigmoid function is not selected; if overfitting is found in the training process, the ReLU function is selected.

[0038] According to the present invention, a temperature monitoring and defect prediction system for additive materials based on the colorimetric temperature measurement principle is provided, comprising:

[0039] Module M1: Conduct additive manufacturing experiments using different process parameters and record the position and temperature signals of the motion system during the printing process;

[0040] Module M2: divide the temperature signal collected according to the colorimetric temperature measurement principle into signal segments with the same time interval and the same length, each signal segment contains the same amount of temperature data, and extract the characteristic parameters of the temperature data in each segment;

[0041] Module M3: Scan the printed additive manufacturing experimental samples to obtain the internal defect information of the samples, combine the position signal data of the motion system, and label each segment of the temperature signal data;

[0042] Module M4: Establish a deep neural network model, divide the data set into training set and test set for training, and save the trained DNN model and parameters; perform laser additive processing, and transfer the position signal and temperature signal of the motion system obtained in real time during the processing to the trained DNN model;

[0043] Module M5: Use the DNN model to automatically monitor the melt pool temperature during the additive process and predict defects.

[0044] Preferably, in the module M1:

[0045] Multiple additive manufacturing experiments were conducted using preset optimal process parameters and manually adjusted secondary process parameters; the process parameters included laser power, scanning speed, and powder feeding amount;

[0046] The details of the additive manufacturing experiment are as follows:

[0047] Explore the optimal process parameters. The optimal process parameters can print components that meet the preset standards. When different process parameters deviate from the values ​​of the optimal process parameters by too much or too little, defects will appear inside the printed components. Design experiments including optimal process parameter combinations and secondary process parameter combinations, covering a variety of situations, build data sets, print single-channel, thin-walled, block or other components according to the designed single-factor experiments, and collect position signals and temperature signals of the motion system during the printing process;

[0048] The temperature signal collected according to the principle of colorimetric temperature measurement is first converted into an analog signal and transmitted to the acquisition card. The acquisition card transmits the analog signal to the PC end, and LabView software is used in the PC end to convert the analog signal into temperature data.

[0049] Preferably, in the module M2:

[0050] The complete temperature data is divided into signal segments with the same time interval and the same length. Adjacent temperature segments contain some of the same temperature data. The characteristic parameters of the temperature data in each segment are extracted, including the highest temperature T in the segment. max , minimum temperature T min , maximum temperature difference ΔT, maximum temperature change rate δT;

[0051] The temperature data characteristic parameters are calculated by the following formula:

[0052] ΔT=T max -T min

[0053]

[0054] Among them, T max is the maximum temperature, T min is the minimum temperature.

[0055] Preferably, in the module M3:

[0056] The printed additive manufacturing experimental samples were subjected to electronic computer tomography to obtain the internal defect information of the samples, and each segment of temperature signal data was labeled in combination with the obtained motion system position signal data;

[0057] Internal defect information includes three categories: pores, cracks, and intact. Labels are manually marked after the additive manufacturing experiment is completed. By obtaining the position signal data and temperature signal data of the motion system during the printing process, the defective parts in the formed parts are matched with the temperature signal data one by one, and the defect conditions of the areas corresponding to different temperature signal sections are manually judged, and one of the three labels of pores, cracks, and intact is marked.

[0058] The labels include a normal defect-free state label, a crack defect state label and a pore defect state label.

[0059] Preferably, in the module M4:

[0060] Establish a deep neural network DNN model, divide the data set into training set and test set, then train it, and save the trained DNN model and parameters;

[0061] The data set includes the input and output of the model. The input is the temperature signal paragraph data, and the output is the label corresponding to the temperature signal paragraph.

[0062] The neural network layers in the DNN model are divided into input layer, hidden layer and output layer. The layers are fully connected. The input layer inputs the highest temperature T max , minimum temperature T min , maximum temperature difference ΔT and maximum temperature change rate δT; the output layer outputs the normal defect-free state label, crack defect state label and pore defect state label; the hidden layer contains multiple layers of neurons, and the number of neurons and layers are adjusted according to the actual data set;

[0063] The input between the fully connected layers passes through the activation function, and the activation function before the output layer is set to SoftMax, and its calculation formula is:

[0064]

[0065] Among them, t represents an array; t i Represents the i-th element of the array; e is a natural constant;

[0066] The activation function of the fully connected layer before the output layer is set to Sigmoid or ReLU function, which is selected and adjusted according to the situation. The calculation formulas are:

[0067]

[0068] ReLU(t)=max(0,t)

[0069] The choice of activation function is adjusted according to the actual operation results and specific loss conditions. When the output result deviates from 0, the Sigmoid function is not selected; if overfitting is found in the training process, the ReLU function is selected.

[0070] Compared with the prior art, the present invention has the following beneficial effects:

[0071] 1. The present invention has low cost and can basically complete the monitoring of the molten pool temperature state and defect prediction of the additive manufacturing process by using only a single colorimetric temperature measurement principle sensor, and the equipment is easy to install;

[0072] 2. The present invention divides the temperature signal into signal segments with the same time length at the same time interval, which can more accurately select the defective part to a certain extent, improve the training accuracy, and obtain a model with better prediction effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0074] Figure 1 is a flow chart of the present invention;

[0075] Figure 2 It is a schematic diagram of temperature section division of the present invention;

[0076] Figure 3 It is a schematic diagram of the neural network structure of the present invention;

[0077] Figure 4 Schematic diagram of the effect of using electronic computed tomography. DETAILED DESCRIPTION

[0078] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0079] Embodiment 1:

[0080] According to the present invention, a method for temperature monitoring and defect prediction of additive materials based on the colorimetric temperature measurement principle is provided. Figure 1-Figure 3 As shown, including:

[0081] Step S1: performing additive manufacturing experiments using different process parameters, and recording the position signal and temperature signal of the motion system during the printing process;

[0082] Specifically, in step S1:

[0083] Multiple additive manufacturing experiments were conducted using preset optimal process parameters and manually adjusted secondary process parameters; the process parameters included laser power, scanning speed, and powder feeding amount;

[0084] The details of the additive manufacturing experiment are as follows:

[0085] Explore the optimal process parameters. The optimal process parameters can print components that meet the preset standards. When different process parameters deviate from the values ​​of the optimal process parameters by too much or too little, defects will appear inside the printed components. Design experiments including optimal process parameter combinations and secondary process parameter combinations, covering a variety of situations, build data sets, print single-channel, thin-walled, block or other components according to the designed single-factor experiments, and collect position signals and temperature signals of the motion system during the printing process;

[0086] The temperature signal collected according to the principle of colorimetric temperature measurement is first converted into an analog signal and transmitted to the acquisition card. The acquisition card transmits the analog signal to the PC end, and LabView software is used in the PC end to convert the analog signal into temperature data.

[0087] Step S2: dividing the temperature signal collected according to the colorimetric temperature measurement principle into signal segments with the same time interval and the same length, each signal segment contains the same amount of temperature data, and extracting characteristic parameters of the temperature data in each segment;

[0088] Specifically, in step S2:

[0089] The complete temperature data is divided into signal segments with the same time interval and the same length. Adjacent temperature segments contain some of the same temperature data. The characteristic parameters of the temperature data in each segment are extracted, including the highest temperature T in the segment. max , minimum temperature T min , maximum temperature difference ΔT, maximum temperature change rate δT;

[0090] The temperature data characteristic parameters are calculated by the following formula:

[0091] ΔT=T max -T min

[0092]

[0093] Among them, T max is the maximum temperature, T min is the minimum temperature.

[0094] Step S3: Scan the printed additive manufacturing experimental sample to obtain the internal defect information of the sample, and label each section of the temperature signal data in combination with the position signal data of the motion system;

[0095] Specifically, in step S3:

[0096] The printed additive manufacturing experimental samples were subjected to electronic computer tomography to obtain the internal defect information of the samples, and each segment of temperature signal data was labeled in combination with the obtained motion system position signal data;

[0097] Internal defect information includes three categories: pores, cracks, and intact. Labels are manually marked after the additive manufacturing experiment is completed. By obtaining the position signal data and temperature signal data of the motion system during the printing process, the defective parts in the formed parts are matched with the temperature signal data one by one, and the defect conditions of the areas corresponding to different temperature signal sections are manually judged, and one of the three labels of pores, cracks, and intact is marked.

[0098] The labels include a normal defect-free state label, a crack defect state label and a pore defect state label.

[0099] Step S4: Establish a deep neural network model, divide the data set into a training set and a test set for training, and save the trained DNN model and parameters; perform laser additive manufacturing, and transfer the position signal and temperature signal of the motion system obtained in real time during the manufacturing process to the trained DNN model;

[0100] Specifically, in step S4:

[0101] Establish a deep neural network DNN model, divide the data set into training set and test set, then train it, and save the trained DNN model and parameters;

[0102] The data set includes the input and output of the model. The input is the temperature signal paragraph data, and the output is the label corresponding to the temperature signal paragraph.

[0103] The neural network layers in the DNN model are divided into input layer, hidden layer and output layer. The layers are fully connected. The input layer inputs the highest temperature T max , minimum temperature T min , maximum temperature difference ΔT and maximum temperature change rate δT; the output layer outputs the normal defect-free state label, crack defect state label and pore defect state label; the hidden layer contains multiple layers of neurons, and the number of neurons and layers are adjusted according to the actual data set;

[0104] The input between the fully connected layers passes through the activation function, and the activation function before the output layer is set to SoftMax, and its calculation formula is:

[0105]

[0106] Among them, t represents an array; t i Represents the i-th element of the array; e is a natural constant, and its value is approximately 2.718;

[0107] The activation function of the fully connected layer before the output layer is set to Sigmoid or ReLU function, which is selected and adjusted according to the situation. The calculation formulas are:

[0108]

[0109] ReLU(t)=max(0,t)

[0110] The choice of activation function is adjusted according to the actual calculation results and specific loss conditions. When the derivative of Sigmoid is near 0, it has better activation. At this time, the Sigmoid function is a better choice. When most of the output results deviate from 0, the Sigmoid function is not selected; the ReLU function is faster to calculate, and can make some neurons output 0, reduce the interdependence of parameters, and reduce the occurrence of overfitting. If the results are found to be overfitting during training, the ReLU function can be selected instead.

[0111] Step S5: Use the DNN model to automatically monitor the temperature of the molten pool during the additive process and predict defects.

[0112] Embodiment 2:

[0113] Embodiment 2 is a preferred example of Embodiment 1, and is used to illustrate the present invention in more detail.

[0114] The present invention also provides an additive temperature monitoring and defect prediction system based on the colorimetric temperature measurement principle. The additive temperature monitoring and defect prediction system based on the colorimetric temperature measurement principle can be realized by executing the process steps of the additive temperature monitoring and defect prediction method based on the colorimetric temperature measurement principle, that is, those skilled in the art can understand the additive temperature monitoring and defect prediction method based on the colorimetric temperature measurement principle as a preferred implementation of the additive temperature monitoring and defect prediction system based on the colorimetric temperature measurement principle.

[0115] According to the present invention, a temperature monitoring and defect prediction system for additive materials based on the colorimetric temperature measurement principle is provided, comprising:

[0116] Module M1: Conduct additive manufacturing experiments using different process parameters and record the position and temperature signals of the motion system during the printing process;

[0117] Specifically, in the module M1:

[0118] Multiple additive manufacturing experiments were conducted using preset optimal process parameters and manually adjusted secondary process parameters; the process parameters included laser power, scanning speed, and powder feeding amount;

[0119] The details of the additive manufacturing experiment are as follows:

[0120] Explore the optimal process parameters. The optimal process parameters can print components that meet the preset standards. When different process parameters deviate from the values ​​of the optimal process parameters by too much or too little, defects will appear inside the printed components. Design experiments including optimal process parameter combinations and secondary process parameter combinations, covering a variety of situations, build data sets, print single-channel, thin-walled, block or other components according to the designed single-factor experiments, and collect position signals and temperature signals of the motion system during the printing process;

[0121] The temperature signal collected according to the principle of colorimetric temperature measurement is first converted into an analog signal and transmitted to the acquisition card. The acquisition card transmits the analog signal to the PC end, and LabView software is used in the PC end to convert the analog signal into temperature data.

[0122] Module M2: divide the temperature signal collected according to the colorimetric temperature measurement principle into signal segments with the same time interval and the same length, each signal segment contains the same amount of temperature data, and extract the characteristic parameters of the temperature data in each segment;

[0123] Specifically, in the module M2:

[0124] The complete temperature data is divided into signal segments with the same time interval and the same length. Adjacent temperature segments contain some of the same temperature data. The characteristic parameters of the temperature data in each segment are extracted, including the highest temperature T in the segment. max , minimum temperature T min , maximum temperature difference ΔT, maximum temperature change rate δT;

[0125] The temperature data characteristic parameters are calculated by the following formula:

[0126] ΔT=T max -T min

[0127]

[0128] Among them, T max is the maximum temperature, T min is the minimum temperature.

[0129] Module M3: Scan the printed additive manufacturing experimental samples to obtain the internal defect information of the samples, combine the position signal data of the motion system, and label each segment of the temperature signal data;

[0130] Specifically, in the module M3:

[0131] The printed additive manufacturing experimental samples were subjected to electronic computer tomography to obtain the internal defect information of the samples, and each segment of temperature signal data was labeled in combination with the obtained motion system position signal data;

[0132] Internal defect information includes three categories: pores, cracks, and intact. Labels are manually marked after the additive manufacturing experiment is completed. By obtaining the position signal data and temperature signal data of the motion system during the printing process, the defective parts in the formed parts are matched with the temperature signal data one by one, and the defect conditions of the areas corresponding to different temperature signal sections are manually judged, and one of the three labels of pores, cracks, and intact is marked.

[0133] The labels include a normal defect-free state label, a crack defect state label and a pore defect state label.

[0134] Module M4: Establish a deep neural network model, divide the data set into training set and test set for training, and save the trained DNN model and parameters; perform laser additive processing, and transfer the position signal and temperature signal of the motion system obtained in real time during the processing to the trained DNN model;

[0135] Specifically, in the module M4:

[0136] Establish a deep neural network DNN model, divide the data set into training set and test set, then train it, and save the trained DNN model and parameters;

[0137] The data set includes the input and output of the model. The input is the temperature signal paragraph data, and the output is the label corresponding to the temperature signal paragraph.

[0138] The neural network layers in the DNN model are divided into input layer, hidden layer and output layer. The layers are fully connected. The input layer inputs the highest temperature T max , minimum temperature T min , maximum temperature difference ΔT and maximum temperature change rate δT; the output layer outputs the normal defect-free state label, crack defect state label and pore defect state label; the hidden layer contains multiple layers of neurons, and the number of neurons and layers are adjusted according to the actual data set;

[0139] The input between the fully connected layers passes through the activation function, and the activation function before the output layer is set to SoftMax, and its calculation formula is:

[0140]

[0141] Among them, t represents an array; t i Represents the i-th element of the array; e is a natural constant;

[0142] The activation function of the fully connected layer before the output layer is set to Sigmoid or ReLU function, which is selected and adjusted according to the situation. The calculation formulas are:

[0143]

[0144] ReLU(t)=max(0,t)

[0145] The choice of activation function is adjusted according to the actual operation results and specific loss conditions. When the output result deviates from 0, the Sigmoid function is not selected; if overfitting is found in the training process, the ReLU function is selected.

[0146] Module M5: Use the DNN model to automatically monitor the melt pool temperature during the additive process and predict defects.

[0147] Embodiment 3:

[0148] Embodiment 3 is a preferred example of Embodiment 1, and is used to illustrate the present invention in more detail.

[0149] like Figure 1 As shown, a method for monitoring the temperature of the molten pool and predicting defects in an additive process based on the principle of colorimetric temperature measurement specifically includes the following steps:

[0150] 1. Several single-channel additive manufacturing experiments were carried out using the optimal process parameters and the artificially adjusted secondary process parameters. The temperature signal in the additive manufacturing process was measured using the colorimetric temperature measurement principle, and the acquisition frequency was set to 60 Hz.

[0151] The specific process of the additive manufacturing experiment is as follows:

[0152] First, we explore the optimal process parameters. The optimal process parameters can print components with smooth and complete surfaces and few internal defects. We design a single-factor experiment with the optimal parameters of laser power 1200W, powder feeding speed 9g / min, and scanning speed 9mm / s as an example, as shown in the following table (including repeated groups, the data in the table are only examples). When different process parameters deviate from the values ​​of the optimal process parameters too much or too little, obvious defects will appear inside the printed components. The designed experiment includes the optimal process parameter combination and the secondary process parameter combination, covering the situations of no defects, few defects, and many defects, so as to facilitate the subsequent construction of a highly versatile data set. Finally, according to the designed single-factor experiment, a single channel, thin wall, block or other components are printed, and the position signal and temperature signal of the motion system during the printing process are collected.

[0153]

[0154] 2. Record the position signal and temperature signal of the motion system in correspondence with each other during the printing process, so as to facilitate the correspondence between the sample defect label and the temperature signal label in the subsequent steps.

[0155] Computer tomography can be used to obtain Figure 4The effect shown is shown in the schematic diagram as an example. The first layer of the framed area is the void defect, the second layer of the framed area is the crack defect, and the third layer of the framed area is the defect-free area. The internal defect information includes three categories: pores, cracks, and no defects.

[0156] The label is a manual mark after the additive manufacturing experiment is completed. By obtaining the position signal data and temperature signal data of the motion system during the printing process, the defective parts in the formed part can be matched with the temperature signal data one by one, and the defect conditions of the areas corresponding to different temperature signal sections can be manually judged, and one of the three marks of pores, cracks, and intact can be marked.

[0157] 3. Segment into signal segments with the same time interval and the same length. Each signal segment is 167ms long. There is a 25ms interval between the beginning of each two signal segments. 40 signal segments are segmented every 1s, and each signal segment contains 10 temperature data. The segmentation method is as follows: Figure 2 shown.

[0158] 4. Use the algorithm to extract the characteristic parameters of 10 temperature data in each signal segment, including the highest temperature T in the segment max , minimum temperature T min , maximum temperature difference ΔT, maximum temperature change rate δT.

[0159] 5. Perform computed tomography (CT) on the printed additive manufacturing single-channel experimental sample to obtain the internal defect information of the sample. Combined with the obtained motion system position signal data, label each temperature signal segment data.

[0160] 6. Establish a deep neural network (DNN) model. DNN network is a deep neural network, such as Figure 3 As shown. The data set is divided into a training set and a test set, and then training is performed. The four temperature data features obtained in step 4 are used as the input of the neural network, and the labels obtained in step 5 are used as the output. Then the hyperparameters of the neural network are optimized and adjusted, and finally the trained DNN model and parameters are saved.

[0161] 7. Perform multiple laser additive manufacturing processes and pass the position and temperature signals of the motion system acquired in real time during the manufacturing process into the trained DNN model.

[0162] 8. Use the obtained DNN model to automatically monitor the melt pool temperature of the additive process and predict defects.

[0163] Laboratory conditions required for training: GTX 3070 graphics card, Windows system, Python programming language, Pytorch deep learning framework.

[0164] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0165] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for additive temperature monitoring and defect prediction based on the principle of colorimetric temperature measurement, characterized in that: include: Step S1: performing additive manufacturing experiments using different process parameters, and recording the position signal and temperature signal of the motion system during the printing process; Step S2: dividing the temperature signal collected according to the colorimetric temperature measurement principle into signal segments with the same time interval and the same length, each signal segment contains the same amount of temperature data, and extracting characteristic parameters of the temperature data in each segment; Step S3: Scan the printed additive manufacturing experimental sample to obtain the internal defect information of the sample, and label each section of the temperature signal data in combination with the position signal data of the motion system; Step S4: Establish a deep neural network model, divide the data set into a training set and a test set for training, and save the trained DNN model and parameters; perform laser additive manufacturing, and transfer the position signal and temperature signal of the motion system obtained in real time during the manufacturing process to the trained DNN model; Step S5: Use the DNN model to automatically monitor the temperature of the molten pool during the additive process and predict defects.

2. The additive material temperature monitoring and defect prediction method based on the colorimetric temperature measurement principle according to claim 1 is characterized in that: In step S1: Multiple additive manufacturing experiments were conducted using preset optimal process parameters and manually adjusted secondary process parameters; the process parameters included laser power, scanning speed, and powder feeding amount; The details of the additive manufacturing experiment are as follows: Explore the optimal process parameters. The optimal process parameters can print components that meet the preset standards. When different process parameters deviate from the values ​​of the optimal process parameters by too much or too little, defects will appear inside the printed components. Design experiments including optimal process parameter combinations and secondary process parameter combinations, covering a variety of situations, build data sets, print single-channel, thin-walled, block or other components according to the designed single-factor experiments, and collect position signals and temperature signals of the motion system during the printing process; The temperature signal collected according to the principle of colorimetric temperature measurement is first converted into an analog signal and transmitted to the acquisition card. The acquisition card transmits the analog signal to the PC end, and LabView software is used in the PC end to convert the analog signal into temperature data.

3. The additive material temperature monitoring and defect prediction method based on the colorimetric temperature measurement principle according to claim 1 is characterized in that: In step S2: The complete temperature data is divided into signal segments with the same time interval and the same length. Adjacent temperature segments contain some of the same temperature data. The characteristic parameters of the temperature data in each segment are extracted, including the highest temperature T in the segment. max , minimum temperature T min , maximum temperature difference ΔT, maximum temperature change rate δT; The temperature data characteristic parameters are calculated by the following formula: ΔT=T max -T min Among them, T max is the maximum temperature, T min is the minimum temperature.

4. The additive material temperature monitoring and defect prediction method based on the colorimetric temperature measurement principle according to claim 1 is characterized in that: In step S3: The printed additive manufacturing experimental samples were subjected to electronic computer tomography to obtain the internal defect information of the samples, and each segment of temperature signal data was labeled in combination with the obtained motion system position signal data; Internal defect information includes three categories: pores, cracks, and intact. Labels are manually marked after the additive manufacturing experiment is completed. By obtaining the position signal data and temperature signal data of the motion system during the printing process, the defective parts in the formed parts are matched with the temperature signal data one by one, and the defect conditions of the areas corresponding to different temperature signal sections are manually judged, and one of the three labels of pores, cracks, and intact is marked. The labels include a normal defect-free state label, a crack defect state label and a pore defect state label.

5. The additive material temperature monitoring and defect prediction method based on the colorimetric temperature measurement principle according to claim 1 is characterized in that: In step S4: Establish a deep neural network DNN model, divide the data set into training set and test set, then train it, and save the trained DNN model and parameters; The data set includes the input and output of the model. The input is the temperature signal paragraph data, and the output is the label corresponding to the temperature signal paragraph. The neural network layers in the DNN model are divided into input layer, hidden layer and output layer. The layers are fully connected. The input layer inputs the highest temperature T max , minimum temperature T min , maximum temperature difference ΔT and maximum temperature change rate δT; the output layer outputs the normal defect-free state label, crack defect state label and pore defect state label; the hidden layer contains multiple layers of neurons, and the number of neurons and layers are adjusted according to the actual data set; The input between the fully connected layers passes through the activation function, and the activation function before the output layer is set to SoftMax, and its calculation formula is: Among them, t represents an array; t i Represents the i-th element of the array; e is a natural constant; The activation function of the fully connected layer before the output layer is set to Sigmoid or ReLU function, which is selected and adjusted according to the situation. The calculation formulas are: ReLU(t)=max(0,t) The choice of activation function is adjusted according to the actual operation results and specific loss conditions. When the output result deviates from 0, the Sigmoid function is not selected; if overfitting is found in the training process, the ReLU function is selected.

6. An additive temperature monitoring and defect prediction system based on the principle of colorimetric temperature measurement, characterized in that: include: Module M1: Conduct additive manufacturing experiments using different process parameters and record the position and temperature signals of the motion system during the printing process; Module M2: divide the temperature signal collected according to the colorimetric temperature measurement principle into signal segments with the same time interval and the same length, each signal segment contains the same amount of temperature data, and extract the characteristic parameters of the temperature data in each segment; Module M3: Scan the printed additive manufacturing experimental samples to obtain the internal defect information of the samples, combine the position signal data of the motion system, and label each segment of the temperature signal data; Module M4: Establish a deep neural network model, divide the data set into training set and test set for training, and save the trained DNN model and parameters; perform laser additive processing, and transfer the position signal and temperature signal of the motion system obtained in real time during the processing to the trained DNN model; Module M5: Use the DNN model to automatically monitor the melt pool temperature during the additive process and predict defects.

7. The additive material temperature monitoring and defect prediction system based on the colorimetric temperature measurement principle according to claim 6 is characterized in that: In the module M1: Multiple additive manufacturing experiments were conducted using preset optimal process parameters and manually adjusted secondary process parameters; the process parameters included laser power, scanning speed, and powder feeding amount; The details of the additive manufacturing experiment are as follows: Explore the optimal process parameters. The optimal process parameters can print components that meet the preset standards. When different process parameters deviate from the values ​​of the optimal process parameters by too much or too little, defects will appear inside the printed components. Design experiments including optimal process parameter combinations and secondary process parameter combinations, covering a variety of situations, build data sets, print single-channel, thin-walled, block or other components according to the designed single-factor experiments, and collect position signals and temperature signals of the motion system during the printing process; The temperature signal collected according to the principle of colorimetric temperature measurement is first converted into an analog signal and transmitted to the acquisition card. The acquisition card transmits the analog signal to the PC end, and LabView software is used in the PC end to convert the analog signal into temperature data.

8. The additive material temperature monitoring and defect prediction system based on the colorimetric temperature measurement principle according to claim 6 is characterized in that: In the module M2: The complete temperature data is divided into signal segments with the same time interval and the same length. Adjacent temperature segments contain some of the same temperature data. The characteristic parameters of the temperature data in each segment are extracted, including the highest temperature T in the segment. max , minimum temperature T min , maximum temperature difference ΔT, maximum temperature change rate δT; The temperature data characteristic parameters are calculated by the following formula: ΔT=T max -T min Among them, T max is the maximum temperature, T min is the minimum temperature.

9. The additive material temperature monitoring and defect prediction system based on the colorimetric temperature measurement principle according to claim 6, characterized in that: In the module M3: The printed additive manufacturing experimental samples were subjected to electronic computer tomography to obtain the internal defect information of the samples, and each segment of temperature signal data was labeled in combination with the obtained motion system position signal data; Internal defect information includes three categories: pores, cracks, and intact. Labels are manually marked after the additive manufacturing experiment is completed. By obtaining the position signal data and temperature signal data of the motion system during the printing process, the defective parts in the formed parts are matched with the temperature signal data one by one, and the defect conditions of the areas corresponding to different temperature signal sections are manually judged, and one of the three labels of pores, cracks, and intact is marked. The labels include a normal defect-free state label, a crack defect state label and a pore defect state label.

10. The additive material temperature monitoring and defect prediction system based on the colorimetric temperature measurement principle according to claim 6, characterized in that: In the module M4: Establish a deep neural network DNN model, divide the data set into training set and test set, then train it, and save the trained DNN model and parameters; The data set includes the input and output of the model. The input is the temperature signal paragraph data, and the output is the label corresponding to the temperature signal paragraph. The neural network layers in the DNN model are divided into input layer, hidden layer and output layer. The layers are fully connected. The input layer inputs the highest temperature T max , minimum temperature T min , maximum temperature difference ΔT and maximum temperature change rate δT; the output layer outputs the normal defect-free state label, crack defect state label and pore defect state label; the hidden layer contains multiple layers of neurons, and the number of neurons and layers are adjusted according to the actual data set; The input between the fully connected layers passes through the activation function, and the activation function before the output layer is set to SoftMax, and its calculation formula is: Among them, t represents an array; t i Represents the i-th element of the array; e is a natural constant; The activation function of the fully connected layer before the output layer is set to Sigmoid or ReLU function, which is selected and adjusted according to the situation. The calculation formulas are: ReLU(t)=max(0,t) The choice of activation function is adjusted according to the actual operation results and specific loss conditions. When the output result deviates from 0, the Sigmoid function is not selected; if overfitting is found in the training process, the ReLU function is selected.

Images