WO2023231244A1 - 考虑焊接温度和顶锻力的fsw抗拉强度智能预测方法 - Google Patents

考虑焊接温度和顶锻力的fsw抗拉强度智能预测方法 Download PDF

Info

Publication number
WO2023231244A1
WO2023231244A1 PCT/CN2022/121213 CN2022121213W WO2023231244A1 WO 2023231244 A1 WO2023231244 A1 WO 2023231244A1 CN 2022121213 W CN2022121213 W CN 2022121213W WO 2023231244 A1 WO2023231244 A1 WO 2023231244A1
Authority
WO
WIPO (PCT)
Prior art keywords
tensile strength
temperature
neural network
convolutional neural
welding
Prior art date
Application number
PCT/CN2022/121213
Other languages
English (en)
French (fr)
Inventor
卢晓红
马冲
贾振元
孙世煊
栾贻函
滕乐
Original Assignee
大连理工大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 大连理工大学 filed Critical 大连理工大学
Publication of WO2023231244A1 publication Critical patent/WO2023231244A1/zh

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
    • B23K20/00—Non-electric welding by applying impact or other pressure, with or without the application of heat, e.g. cladding or plating
    • B23K20/12—Non-electric welding by applying impact or other pressure, with or without the application of heat, e.g. cladding or plating the heat being generated by friction; Friction welding
    • B23K20/122—Non-electric welding by applying impact or other pressure, with or without the application of heat, e.g. cladding or plating the heat being generated by friction; Friction welding using a non-consumable tool, e.g. friction stir welding
    • B23K20/123—Controlling or monitoring the welding process
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
    • B23K20/00—Non-electric welding by applying impact or other pressure, with or without the application of heat, e.g. cladding or plating
    • B23K20/26—Auxiliary equipment
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
    • B23K31/00—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups
    • B23K31/006—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups relating to using of neural networks
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
    • B23K31/00—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups
    • B23K31/02—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups relating to soldering or welding
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
    • B23K31/00—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups
    • B23K31/12—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups relating to investigating the properties, e.g. the weldability, of materials
    • B23K31/125—Weld quality monitoring
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J5/00—Radiation pyrometry, e.g. infrared or optical thermometry
    • G01J5/48—Thermography; Techniques using wholly visual means
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J5/00—Radiation pyrometry, e.g. infrared or optical thermometry
    • G01J2005/0077—Imaging
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Definitions

  • the invention belongs to the field of friction stir welding welding quality prediction, and specifically relates to an intelligent prediction method for FSW tensile strength that takes into account the welding temperature and upsetting force. It uses an infrared thermal imager and a force measuring instrument to collect data, and uses a one-dimensional convolutional neural network to collect data. Deep learning model for tensile strength prediction.
  • the friction stir welding head consists of a shaft shoulder and a stirring needle.
  • the shoulder provides heat through friction during the friction stir welding process, softening the metal to be welded and preventing softening.
  • the final metal flows out of the welding area; the stirring needle rotates and stirs in the welding area to enhance the fluidity of the softened metal.
  • Friction stir welding technology is widely used in aerospace, shipbuilding, automobile and other fields. Tensile strength is usually used as the evaluation standard for welding quality. According to different product needs, the tensile strength must be qualified before it can be applied to production.
  • Luigi et al. used ANN neural network in "Prediction of the Vickers Microhardness and Ultimate Tensile Strength of AA5754 H111 Friction Stir Welding Butt Joints Using Artificial Neural Network" to use the welding parameters during the welding process, the thermal data during the friction stir welding process and the sample position As input to predict AA5754 H111FSW microhardness and tensile strength values.
  • GPR Gaussian regression
  • SVM support vector machine
  • Correlation coefficient (CC) and root mean square error (RMSE) are used as prediction accuracy evaluation parameters and it is concluded that artificial neural network is the most effective in predicting tensile strength.
  • Fan Wenxue et al. used response surface method to study the welding process parameters, including welding speed, stirring head rotation speed and shoulder indentation depth, on the weld resistance Effect of tensile strength.
  • Friction stir welding process parameters include stir head rotation speed, welding speed, etc. Different parameter combinations can form welds of different qualities, and the corresponding microstructure and mechanical properties are also different.
  • the welding process parameters affect the tensile strength by affecting the upsetting force and temperature. The tensile strength changes with changes in upsetting force and welding temperature.
  • Existing studies predicting tensile strength through welding process parameters cannot accurately reflect the dynamic impact of welding process parameters on tensile strength during the welding process. At the same time, existing studies do not The effects of upsetting force and temperature time series data on tensile strength are comprehensively considered.
  • the technical problem to be solved by this invention is to overcome the shortcomings of the existing technology, comprehensively consider the upsetting force that directly affects the tensile strength and the temperature of the characteristic points on the advancing side and the retreating side, and propose a one-dimensional convolutional neural network to use friction stir welding.
  • the temperature and upsetting force time series data of the characteristic points on the advancing and retreating sides of the stir head shoulder on the surface of the weldment are used as input to predict the tensile strength, providing a reference for welding process control.
  • An intelligent prediction method for FSW tensile strength considering welding temperature and upsetting force Infrared thermal imaging camera and force measuring instrument are used to obtain the temperature of the characteristic points on the advancing and retreating sides of the shoulder periphery of the weldment surface and the upsetting force time during the welding process. Sequence data, using the acquired temperature data and upsetting force data to train and test the one-dimensional convolutional neural network, and achieve the tensile strength prediction of friction stir welding. The specific steps are as follows:
  • Step 1 During the friction stir welding process, obtain the temperature of the characteristic point temperature on the advancing side, the characteristic point temperature on the retreating side and the upsetting force data on the periphery of the shoulder on the surface of the weldment during the welding process through an infrared thermal imager and a force measuring instrument;
  • Step 2 The sampling frequencies of the infrared thermal imaging camera and the force measuring instrument are different.
  • the common multiple of the sampling frequencies of the infrared thermal imaging camera and the force measuring instrument is used as the common sampling frequency fs.
  • the sampling frequencies of the two devices are controlled to be the same to obtain the forward side. Characteristic point temperature, retreating side characteristic point temperature and upsetting force time series data, while determining the number of sampling points N according to the welding duration;
  • Step 3 Calculate the length of the welding interval based on the welding speed, sampling frequency and number of sampling points, and obtain the tensile strength of this interval through a tensile test;
  • Step 4 Process the temperature of the forward and backward side characteristic points and the upsetting force data processed in step 2 into an N*3 matrix, process the tensile strength obtained in step 3 into a 1*1 matrix, and create a data set for a Training and testing of dimensional convolutional neural networks;
  • Step 5 Establish a one-dimensional convolutional neural network: There are three inputs to the one-dimensional convolutional neural network, which are the characteristic point temperature on the advancing side, the characteristic point temperature on the retreating side, and the upsetting force time series, and the output is the tensile strength value; According to step 4, the input matrix of the one-dimensional convolutional neural network is N*3 and the output matrix is 1*1.
  • Step 6 Use the Huber robust loss function as the evaluation criterion, use the N*3 temperature and upsetting force matrix processed in step 4 as the input, and the 1*1 tensile strength matrix as the output to train the model, and optimize the model through the Huber loss function. Optimize the parameters in step 5, and finally save the one-dimensional convolutional neural network prediction model under the parameter combination with the highest prediction accuracy;
  • Step 7 During the actual welding process, perform steps 2 and 4 on the temperature of the forward side characteristic point, the temperature of the characteristic point on the retreat side and the upsetting force data on the periphery of the shoulder on the surface of the weldment obtained by the infrared thermal imaging camera and the force measuring instrument. , and then used as the input of the optimized one-dimensional convolutional neural network model in step 6 to predict the tensile strength during the actual processing process.
  • Beneficial effects of the present invention Through a method that combines experiment and theory, a one-dimensional convolutional neural network is established that uses the upsetting force during the friction stir welding process and the temperature time series data of the forward and backward characteristic points on the peripheral shoulder of the weldment surface as input.
  • the network tensile strength prediction model finally achieves the prediction value of tensile strength without conducting destructive tests on weldments.
  • Figure 1 is a one-dimensional convolutional neural network structure diagram
  • Figure 2 is the loss function of the training set and validation set
  • Figure 3 is the tensile strength prediction result based on one-dimensional convolutional neural network.
  • This example uses an infrared thermal imager, a force measuring instrument and a one-dimensional convolutional neural network based on Python to predict the tensile strength of 2219 aluminum alloy friction stir welding.
  • the specific steps are as follows:
  • Step 1 During the friction stir welding experiment, the temperature is measured with an infrared thermal imager, and the temperature of the characteristic points on the forward and backward sides of the shoulder periphery of the weldment surface is collected, the upsetting force data is obtained through the force measuring instrument, and the tensile strength is obtained through the tensile test. value;
  • Step 2 Use the resample function in the pandas class in python to resample the temperature data, and conduct a tensile strength test based on the sampling frequency, welding speed and number of sampling points to obtain the tensile strength value. Correspond the data of upsetting force, advancing side temperature, retreating side temperature and tensile strength value one by one.
  • the format is as shown in Table 1. Table 1 only contains one set of training data. There are actually 50 sets of data obtained from the experiment (30 sets are used for training, 10 sets are used for verification, and 10 sets are used for testing);
  • Step 3 Write a functional function in python, first read the data set obtained in step 2, and normalize the input quantities in the training set, that is, the upsetting force, the forward side feature point temperature, and the retreat side feature point temperature data. . Then the data set is segmented according to the input of the one-dimensional convolutional neural network which is the 30*3 advancing and retreating side temperature and upsetting force matrix, and the output is the 1*1 tensile strength matrix;
  • Step 4 Import the keras library in python, write a one-dimensional convolutional neural network, take 30*3 upsetting force, forward side feature point temperature, and retreat side feature point temperature matrix as input, perform convolution operation through Conv1, and then perform Batch normalization, then convolution through Conv2, pooling through pooling layer 1, then second batch normalization, convolution through Conv3, and pooling through pooling layer 2 , and finally through the fully connected layer 1 and the fully connected layer 2, the tensile strength value is output at the output layer.
  • the number of convolutional layers is 3
  • the number and size of convolution kernels are 265 ⁇ 5, 512 ⁇ 6, and 1024 ⁇ 6 respectively
  • the number of pooling layers is 2
  • the pooling layer size is 3, fully connected.
  • the number of layers is 2, the number of units is 30 and 50 respectively, and the activation function uses ReLU;
  • Step 6 Call the trained neural network model to predict the test set and save the prediction results.
  • One-dimensional convolutional neural network is used to predict the tensile strength value of 2219 aluminum alloy. See Figure 3. From the prediction results, it can be seen that the average relative error between the predicted tensile strength value and the experimental value is about 2%. It can be proved that using this The method can effectively predict tensile strength.

Abstract

一种考虑焊接温度和顶锻力的FSW抗拉强度智能预测方法，运用实验和理论结合的方法，进行搅拌摩擦焊实验，采用红外热像仪和测力仪获取焊接过程中焊件表面轴肩外围前进侧和后退侧特征点温度、顶锻力时间序列数据，利用获取的温度数据和顶锻力数据对一维卷积神经网络进行训练及测试，实现搅拌摩擦焊的抗拉强度预测，为焊接过程调控提供参考。

Description

考虑焊接温度和顶锻力的FSW抗拉强度智能预测方法 技术领域
本发明属于搅拌摩擦焊焊接质量预测领域，具体涉及一种考虑焊接温度和顶锻力的FSW抗拉强度智能预测方法，使用红外热成像仪、测力仪采集数据，通过一维卷积神经网络深度学习模型进行抗拉强度预测。
背景技术
随着航空航天工业的快速发展，航天需求对航天器制造技术有了更高的要求。1991年英国焊接研究所发明了搅拌摩擦焊技术(FSW)，搅拌摩擦焊搅拌头由轴肩和搅拌针组成，轴肩在搅拌摩擦焊接过程中通过摩擦提供热量，使被焊金属软化并且阻止软化后的金属流出焊接区域；搅拌针在焊接区域进行旋转搅拌增强软化金属的流动性。搅拌摩擦焊产生的较低热输入量有效地解决了铝合金和镁合金等低熔点合金在传统焊接中存在的板材变形较大、焊接速度慢等问题。搅拌摩擦焊技术广泛应用于航空航天、造船、汽车等领域，通常采用抗拉强度作为焊接质量的评价标准，根据不同的产品需要，抗拉强度必须合格才能应用到生产中。
Luigi等在《Prediction of the Vickers Microhardness and Ultimate Tensile Strength of AA5754 H111 Friction Stir Welding Butt Joints Using Artificial Neural Network》中采用ANN神经网络，以焊接过程中的焊接参数、搅拌摩擦焊过程中热数据以及样品位置为输入来预测AA5754 H111FSW显微硬度以及抗拉强度值。Mohammad W.Dewan在《Prediction of tensile strength of friction stir weld joints with adaptive neuro-fuzzy inference system(ANFIS)and neural network》中建立了一种优化的自适应神经模糊推理系统(ANFIS)模型，以转速、焊接速度和顶锻力三个焊接参数为输入预测AA2219-T87焊缝抗拉强度。Verma在《Prediction of  tensile behavior of FS welded AA7039 using machine learning》中等采用高斯回归(GPR)、支持向量机(SVM)、人工神经网络和线性回归等机器学习的方法以转速、焊接速度和倾斜角度为输入预测AA7039铝合金搅拌摩擦焊极限抗拉强度，采用相关系数(CC)和均方根误差(RMSE)作为预测精度评估参数并得出结论，人工神经网络对抗拉强度预测最有效。范文学等人在《基于响应面法7A52高强铝合金FSW接头抗拉强度预测及优化》中采用响应曲面法研究了焊接工艺参数，包括焊接速度和搅拌头转速及轴肩压深对焊缝抗拉强度的影响。
搅拌摩擦焊工艺参数包括搅拌头转速、焊接速度等，不同的参数组合能够形成不同质量的焊缝，相应的微观组织和力学性能也有所差异。而焊接工艺参数通过影响顶锻力和温度来影响抗拉强度。抗拉强度随着顶锻力和焊接温度的变化而变化，现有研究通过焊接工艺参数对抗拉强度进行预测并不能准确反映焊接过程中焊接工艺参数对抗拉强度动态影响，同时，现有研究没有综合考虑顶锻力和温度时间序列数据对抗拉强度影响。
发明内容
本发明所要解决的技术问题是克服现有技术的不足，综合考虑直接影响抗拉强度的顶锻力和前进侧、后退侧特征点温度，提出一种一维卷积神经网络，以搅拌摩擦焊过程中焊件表面搅拌头轴肩外围前进侧和后退侧特征点温度、顶锻力时间序列数据为输入实现抗拉强度预测，为焊接过程调控提供参考。
本发明的技术方案：
一种考虑焊接温度和顶锻力的FSW抗拉强度智能预测方法，采用红外热像仪和测力仪获取焊接过程中焊件表面轴肩外围前进侧和后退侧特征点温度、顶锻力时间序列数据，利用获取的温度数据和顶锻力数据对一维卷积神经网络进行训练及测试，实现了搅拌摩擦焊的抗拉强度预测，具体步骤如下：
步骤1：搅拌摩擦焊过程中，通过红外热像仪和测力仪获取焊接过程中焊件表面轴肩外围前进侧特征点温度、后退侧特征点温度和顶锻力数据；
步骤2：红外热像仪和测力仪采样频率不同，将红外热像仪和测力仪采样频率公倍数作为共同采样频率fs，通过重采样的方法，控制两种设备采样频率相同，获得前进侧特征点温度、后退侧特征点温度以及顶锻力时间序列数据，同时根据焊接时长确定采样点数量N；
步骤3：根据焊接速度和采样频率以及采样点数量计算焊接区间长度，通过拉伸试验获取此区间抗拉强度；
步骤4：将步骤2处理后的前进及后退侧特征点温度、顶锻力数据处理成N*3矩阵，将步骤3获取的抗拉强度处理成1*1矩阵，制作成数据集用于一维卷积神经网络的训练和测试；
步骤5：建立一维卷积神经网络：一维卷积神经网络输入量有三个，分别为前进侧特征点温度、后退侧特征点温度以及顶锻力时间序列，输出量为抗拉强度值；根据步骤4，一维卷积神经网络输入矩阵为N*3，输出矩阵为1*1，建立一维卷积神经网络，预设一维卷积神经网络的参数，包括卷积层层数、卷积核数量及尺寸、池化层层数及大小、全连接层层数及单元数以及激活函数；
步骤6：以Huber鲁棒损失函数为评价标准，以步骤4中处理后N*3温度及顶锻力矩阵为输入，1*1抗拉强度矩阵为输出对模型进行训练，通过Huber损失函数优劣对步骤5中的参数进行优化，最终保存预测精度最高的参数组合下的一维卷积神经网络预测模型；
步骤7：在实际焊接加工过程中对红外热像仪和测力仪获取的焊件表面轴肩外围前进侧特征点温度、后退侧特征点温度和顶锻力数据进行步骤2和步骤4的处理，然后作为步骤6中优化后的一维卷积神经网络模型的输入对实际加工 过程中的抗拉强度进行预测。
本发明的有益效果：通过实验与理论相结合的方法，建立以搅拌摩擦焊接过程中顶锻力和焊件表面轴肩外围前进及后退侧特征点温度时间序列数据作为输入的一维卷积神经网络抗拉强度预测模型，最终实现在不对焊件进行破坏性试验前提下，获得抗拉强度预测值。
附图说明
图1是一维卷积神经网络结构图；
图2是训练集和验证集的损失函数；
图3是基于一维卷积神经网络的抗拉强度预测结果。
具体实施方式
下面结合技术方案和附图详细说明本发明的具体实施方式，但本发明不受实施例的限制。
本实施例使用红外热像仪、测力仪和基于python建立的一维卷积神经网络，对2219铝合金搅拌摩擦焊抗拉强度进行预测，具体步骤如下：
步骤1：搅拌摩擦焊实验过程中通过红外热像仪测温，采集焊件表面轴肩外围前进及后退侧特征点温度，通过测力仪获得顶锻力数据，通过拉伸试验获得抗拉强度值；
步骤2：使用python中pandas类中的resample函数对温度数据进行重采样处理，同时结合采样频率、焊接速度以及采样点数量进行抗拉强度试验获得抗拉强度值。将顶锻力、前进侧温度、后退侧温度及抗拉强度值数据一一对应。格式如表1，表1中只包含一组训练数据，实际从实验中获取的数据共有50组(30组用作训练，10组用作验证，10组用作测试)；
表1
Figure PCTCN2022121213-appb-000001
步骤3：通过python编写功能函数，首先读取步骤2中获得的数据集，并对训练集中的输入量，即顶锻力、前进侧特征点温度、后退侧特征点温度数据进行归一化处理。然后根据一维卷积神经网络输入为30*3的前进、后退侧温度和顶锻力矩阵，输出为1*1抗拉强度矩阵对数据集进行分段；
步骤4：在python中导入keras库，编写一维卷积神经网络，以30*3顶锻力、前进侧特征点温度、后退侧特征点温度矩阵为输入，经过Conv1进行卷积 运算，然后进行批标准化处理，然后经过Conv2进行卷积运算，再经过池化层1进行池化运算，然后进行第二次批标准化处理，再通过Conv3进行卷积运算，然后通过池化层2进行池化运算，最后经过全连接层1和全连接层2，在输出层输出抗拉强度值。参见图1，卷积层层数为3，卷积核数量及尺寸分别为265×5、512×6以及1024×6，池化层层数为2，池化层大小均为3、全连接层层数为2，单元数分别为30和50，激活函数采用ReLU；
步骤5：使用建立好的一维卷积神经网络模型并将数据集按照训练集：验证集：测试集＝6：2：2划分，进行500轮训练，初始学习率设为0.1、绘制误差曲线并且保存完成训练的神经网络模型，参见图2；
步骤6：调用已训练好的神经网络模型，对测试集进行预测并保存预测结果。
利用一维卷积神经网络预测2219铝合金抗拉强度值，参见图3，从预测结果可以看出，抗拉强度预测值与实验值平均相对误差在百分之二左右，可以证明，使用该方法可以对抗拉强度进行有效预测。

Claims (1)

  1. 一种考虑焊接温度和顶锻力的FSW抗拉强度智能预测方法，其特征在于，采用红外热像仪和测力仪获取焊接过程中焊件表面轴肩外围前进侧和后退侧特征点温度、顶锻力时间序列数据，利用获取的温度数据和顶锻力数据对一维卷积神经网络进行训练及测试，实现搅拌摩擦焊的抗拉强度预测，具体步骤如下：
    步骤1：搅拌摩擦焊过程中，通过红外热像仪和测力仪获取焊接过程中焊件表面轴肩外围前进侧特征点温度、后退侧特征点温度和顶锻力数据；
    步骤2：红外热像仪和测力仪采样频率不同，将红外热像仪和测力仪采样频率公倍数作为共同采样频率fs，通过重采样的方法，控制两种设备采样频率相同，获得前进侧特征点温度、后退侧特征点温度以及顶锻力时间序列数据，同时根据焊接时长确定采样点数量N；
    步骤3：根据焊接速度和采样频率以及采样点数量计算焊接区间长度，通过拉伸试验获取此区间抗拉强度；
    步骤4：将步骤2处理后的前进及后退侧特征点温度、顶锻力数据处理成N*3矩阵，将步骤3获取的抗拉强度处理成1*1矩阵，制作成数据集用于一维卷积神经网络的训练和测试；
    步骤5：建立一维卷积神经网络：一维卷积神经网络输入量有三个，分别为前进侧特征点温度、后退侧特征点温度以及顶锻力时间序列，输出量为抗拉强度值；根据步骤4，一维卷积神经网络输入矩阵为N*3，输出矩阵为1*1，建立一维卷积神经网络，预设一维卷积神经网络的参数，包括卷积层层数、卷积核数量及尺寸、池化层层数及大小、全连接层层数及单元数以及激活函数；
    步骤6：以Huber鲁棒损失函数为评价标准，以步骤4中处理后N*3温度及顶锻力矩阵为输入，1*1抗拉强度矩阵为输出对模型进行训练，通过Huber损失函数优劣对步骤5中的参数进行优化，最终保存预测精度最高的参数组合 下的一维卷积神经网络预测模型；
    步骤7：在实际焊接加工过程中对红外热像仪和测力仪获取的焊件表面轴肩外围前进侧特征点温度、后退侧特征点温度和顶锻力数据进行步骤2和步骤4的处理，然后作为步骤6中优化后的一维卷积神经网络模型的输入对实际加工过程中的抗拉强度进行预测。
PCT/CN2022/121213 2022-06-02 2022-09-26 考虑焊接温度和顶锻力的fsw抗拉强度智能预测方法 WO2023231244A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210625492.8A CN114888426B (zh) 2022-06-02 2022-06-02 考虑焊接温度和顶锻力的fsw抗拉强度智能预测方法
CN202210625492.8 2022-06-02

Publications (1)

Publication Number Publication Date
WO2023231244A1 true WO2023231244A1 (zh) 2023-12-07

Family

ID=82725790

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/121213 WO2023231244A1 (zh) 2022-06-02 2022-09-26 考虑焊接温度和顶锻力的fsw抗拉强度智能预测方法

Country Status (2)

Country Link
CN (1) CN114888426B (zh)
WO (1) WO2023231244A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117697119A (zh) * 2024-02-04 2024-03-15 上海耀焊科技有限公司 一种惯性摩擦焊机智能监测控制方法和设备

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114888426B (zh) * 2022-06-02 2023-02-14 大连理工大学 考虑焊接温度和顶锻力的fsw抗拉强度智能预测方法
CN116140786B (zh) * 2023-03-06 2023-07-14 四川艾庞机械科技有限公司 搅拌摩擦焊接方法及其系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010050834A1 (en) * 2008-11-03 2010-05-06 Instituto De Soldadura E Qualidade System and process for automatic determination of welding parameters for automated friction stir welding
CN109977948A (zh) * 2019-03-20 2019-07-05 哈尔滨工业大学 一种基于卷积神经网络的搅拌摩擦焊缝缺陷识别方法
WO2021007812A1 (zh) * 2019-07-17 2021-01-21 深圳大学 一种深度神经网络超参数优化方法、电子设备及存储介质
CN112388148A (zh) * 2020-09-25 2021-02-23 广州瑞松智能科技股份有限公司 基于模拟退火算法的焊接参数优化法
CN112643190A (zh) * 2021-01-21 2021-04-13 中车长春轨道客车股份有限公司 基于聚类算法的搅拌摩擦焊质量评价方法、装置及系统
CN114888426A (zh) * 2022-06-02 2022-08-12 大连理工大学 考虑焊接温度和顶端力的fsw抗拉强度智能预测方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103212783B (zh) * 2013-04-28 2015-12-23 江苏科技大学 一种搅拌摩擦焊温度场测量专用垫板
CN109590598B (zh) * 2018-12-13 2020-05-01 山东大学 一种摩擦预热的惯性摩擦焊接工艺方法
CN114289855B (zh) * 2022-01-21 2023-08-22 沈阳航空航天大学 一种改善搅拌摩擦焊缝组织不对称性的焊接方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010050834A1 (en) * 2008-11-03 2010-05-06 Instituto De Soldadura E Qualidade System and process for automatic determination of welding parameters for automated friction stir welding
CN109977948A (zh) * 2019-03-20 2019-07-05 哈尔滨工业大学 一种基于卷积神经网络的搅拌摩擦焊缝缺陷识别方法
WO2021007812A1 (zh) * 2019-07-17 2021-01-21 深圳大学 一种深度神经网络超参数优化方法、电子设备及存储介质
CN112388148A (zh) * 2020-09-25 2021-02-23 广州瑞松智能科技股份有限公司 基于模拟退火算法的焊接参数优化法
CN112643190A (zh) * 2021-01-21 2021-04-13 中车长春轨道客车股份有限公司 基于聚类算法的搅拌摩擦焊质量评价方法、装置及系统
CN114888426A (zh) * 2022-06-02 2022-08-12 大连理工大学 考虑焊接温度和顶端力的fsw抗拉强度智能预测方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117697119A (zh) * 2024-02-04 2024-03-15 上海耀焊科技有限公司 一种惯性摩擦焊机智能监测控制方法和设备
CN117697119B (zh) * 2024-02-04 2024-05-03 上海耀焊科技有限公司 一种惯性摩擦焊机智能监测控制方法和设备

Also Published As

Publication number Publication date
CN114888426B (zh) 2023-02-14
CN114888426A (zh) 2022-08-12

Similar Documents

Publication Publication Date Title
WO2023231244A1 (zh) 考虑焊接温度和顶锻力的fsw抗拉强度智能预测方法
Zhou et al. Overview of recent advances of process analysis and quality control in resistance spot welding
Sudhagar et al. Monitoring of friction stir welding based on vision system coupled with Machine learning algorithm
Du et al. Conditions for void formation in friction stir welding from machine learning
Gisario et al. Springback control in sheet metal bending by laser-assisted bending: Experimental analysis, empirical and neural network modelling
Sahu et al. Hybrid fuzzy-grey-Taguchi based multi weld quality optimization of Al/Cu dissimilar friction stir welded joints
Geng et al. Numerical and experimental investigation on friction welding of austenite stainless steel and middle carbon steel
DebRoy et al. Tool durability maps for friction stir welding of an aluminium alloy
Chadha et al. A survey of machine learning in friction stir welding, including unresolved issues and future research directions
Zhao et al. Quality prediction and rivet/die selection for SPR joints with artificial neural network and genetic algorithm
Zhao et al. Modelling and interaction analysis of the self-pierce riveting process using regression analysis and FEA
Raj et al. Weld quality monitoring via machine learning-enabled approaches
Kalita et al. Finite element modelling, predictive modelling and optimization of metal inert gas, tungsten inert gas and friction stir welding processes: a comprehensive review
CN111537368A (zh) 一种适用于搅拌摩擦焊接头的变幅多级加载疲劳寿命预测方法
Chanakyan et al. Friction stir processing (FSP) of numerical study based on design of experiment-review
Ye et al. Hybrid modeling of mechanical properties and hardness of aluminum alloy 5083 and C100 Copper with various machine learning algorithms in friction stir welding
Patel et al. CAFE modeling, neural network modeling, and experimental investigation of friction stir welding
Yang et al. Computational fluid dynamics model of AA6061 friction stir welding with considering mechanical anisotropy
Akbari et al. Simulation of dissimilar friction stir welding of AA7075 and AA5083 aluminium alloys using Coupled Eulerian–Lagrangian approach
Dorbane et al. A tree-driven ensemble learning approach to predict FS welded Al-6061-T6 material behavior
Kolahgar et al. Formability of friction stir-welded blanks with different thickness ratios
Sivam et al. Multi response optimization of setting input variables for getting better cylindrical cups in sheet metal spinning of Al 6061–T6 by Grey relation analysis
Park et al. Control of the weld quality using welding parameters in a robotic welding process
Maheshwari et al. Deep Neural Network-Based Approach for Modeling, Predicting, and Validating Weld Quality and Mechanical Properties of Friction Stir Welded Dissimilar Materials
Arya et al. Study on Friction Stir Welding of Aluminium Plates Using an Artificial Neural Network

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22944553

Country of ref document: EP

Kind code of ref document: A1