A variable environment thermal wave hole detection method based on physical and data double driving

Abstract

The application discloses a variable environment thermal wave hole detection method based on physical and data double driving, relates to a variable environment thermal wave hole detection method, and realizes final detection by combining a physical model and a digital model; the physical model is composed of finite element simulation simulation, a thermal imager and the Internet; the digital model is composed of a convolutional neural network and a support vector machine; original data are obtained by using the physical model, are transmitted into the digital model through the Internet for training and detection, and finally, the detection of the variable environment thermal wave hole based on physical and data double driving is realized; the method identifies real detection environment parameters in reverse, avoids the problem of inaccurate detection caused by environment change, can rapidly and accurately detect the material internal hole under the detection environment, and has high application value and conversion prospect.

Description

Technical Field

[0001] This invention relates to a method for detecting thermal wave holes in varying environments, and more particularly to a method for detecting thermal wave holes in varying environments based on a dual-drive approach of physics and data. Background Technology

[0002] Currently, there are many problems with infrared thermography detection. Common methods for detecting hole defects can only measure the size of the hole area; there is a lack of effective methods for accurately measuring the hole depth.

[0003] For infrared thermal wave detection, changes in the experimental environment can lead to increased detection errors. Therefore, there is an urgent need for a reverse analysis method driven by both physics and data to quickly and accurately identify the actual detection environment. Summary of the Invention

[0004] The purpose of this invention is to provide a method for detecting thermal wave holes in varying environments based on both physical and data-driven approaches. This invention is an efficient, intelligent, and convenient method for detecting thermal wave holes in varying environments based on both physical and data-driven approaches, which can more intelligently identify the actual detection environment.

[0005] The method proposed in this invention is achieved through the following technical solution: A method for detecting thermal wave holes in varying environments based on a dual-drive approach of physics and data is proposed. The method combines a physical model and a digital model to achieve the final detection. The physical model consists of finite element simulation, a thermal imager, and the Internet. The digital model consists of a convolutional neural network and a support vector machine. The raw data is obtained using the physical model and transmitted to the digital model via the Internet for training and detection, ultimately achieving dual-drive detection of thermal wave holes in varying environments using both physics and data. Includes the following steps: The environmental parameters were set using a finite element simulation program to establish a finite element heat transfer model. Then, the simulation model images were extracted and transmitted via the Internet. In the actual experiment, the size, location, and depth of the holes were measured. Thermal images of the site were collected using a thermal imager and then transmitted via the Internet. The data generated by the heat transfer model established using the finite element method and the thermal images of the site collected by the thermal imager were transmitted to a convolutional neural network for training and detection to obtain the location, area, and time of hole appearance. The detection results of the heat transfer model data established by finite element method are compared with the detection results of thermal imager experimental data to check whether the hole position error, area error, and occurrence time difference are less than e1, e2, and e3, respectively. If the result is "yes", the convective heat dissipation coefficient and heating power of the on-site environment are obtained; if the result is "no", the environmental coefficient and heating power are updated through Monto Carlo, the finite element heat transfer model is re-established, and the experiment is carried out in sequence. By using a convolutional neural network to obtain the location, size, and time of hole appearance, and by comparing the convective heat dissipation coefficient and heating power of the on-site environment obtained through finite element heat transfer model experiments and thermal images, and by using an SVM support vector machine to extract features for training and prediction, the information of hole size, location, and depth is finally obtained.

[0006] The advantages and effects of this invention are: (1) This invention achieves the final detection effect by combining a physical model and a digital model. The physical model consists of finite element simulation, a thermal imager, and the Internet. The digital model consists of a convolutional neural network and a support vector machine. Raw data is obtained using the physical model and transmitted to the digital model via the Internet for training and detection, ultimately realizing the detection of thermal wave holes in changing environments through a dual-drive approach of physical and data analysis.

[0007] (2) This invention combines infrared thermal wave detection with deep learning networks, using convolutional neural networks to detect the location coordinates, area size and detection time of holes, and then using SVM support vector machine to extract features and predict hole depth.

[0008] (3) Compare the detection results of the heat transfer model data established by the finite element method with the detection results of the thermal imager experiment data to check whether the hole position error, area error, and occurrence time difference are less than e1, e2, and e3, respectively. If the result is "yes", the convective heat dissipation coefficient and heating power of the on-site environment are obtained. If the result is "no", the environmental coefficient and heating power are updated through Monto Carlo, the finite element heat transfer model is re-established, and the experiment is carried out in sequence. This method can obtain the temperature, power, and other parameters of the real experimental environment through reverse design, avoiding experimental changes and inaccuracies caused by slight changes in the environment.

[0009] (4) The requirements for equipment purchase and installation are low, saving time, labor and financial costs. Attached Figure Description

[0010] Figure 1 This is a flowchart of the method and apparatus for detecting thermal wave holes in varying environments based on both physical and data-driven approaches according to the present invention. Implementation

[0011] The following is an appendix to the instruction manual. Figure 1 The present invention will be described in detail below.

[0012] This invention discloses a method for detecting thermal pores in varying environments based on a dual-drive approach of physics and data. This detection method and device combine a physical model and a digital model to achieve the final detection effect. The specific steps are as follows: Environmental parameters are set using Workbench to establish a finite element heat transfer model, followed by extracting images of the simulation model and transmitting the data over the internet. Then, a thermal imager is used to acquire thermal images of the site, which are then transmitted over the internet. The data generated by the finite element heat transfer model and the thermal images acquired by the thermal imager are transmitted to a convolutional neural network for training and detection to obtain the pore location, pore area, and pore appearance time. The detection results from the two experiments are compared to examine the pore location error, area error, and appearance time difference, determining whether it is necessary to update the finite element heat transfer model in reverse to obtain the convective heat dissipation coefficient and heating power of the site environment. Subsequently, the location, size, and time of hole appearance were obtained by using a convolutional neural network, along with the convective heat dissipation coefficient and heating power of the on-site environment obtained by comparing finite element heat transfer model experiments and thermal images. These data were then trained and tested using a support vector machine to ultimately obtain information on the size, location, and depth of the hole.

[0013] like Figure 1 As shown in the figure, this is a flowchart of the method and apparatus for detecting thermal wave holes in varying environments based on both physical and data-driven approaches according to the present invention. The implementation method includes the following steps: (1) First, select whether to perform training or testing on the device. If using a training model, use Workbench to set the environmental parameters, establish a finite element heat transfer model, then extract the simulation model image and transmit the data over the Internet. Next, conduct the actual experiment, measuring the size, location, and depth of the holes. Use a thermal imager to collect thermal images from the site, and then transmit the data over the Internet. Transmit the data generated by the finite element heat transfer model and the thermal images collected by the thermal imager to a convolutional neural network for training and testing to obtain the hole location, hole area, and the time of hole appearance.

[0014] (2) If a detection model is used, a thermal imager is used to collect thermal images of the scene, and then the data is transmitted via the Internet. The thermal images of the scene collected by the thermal imager are transmitted to a convolutional neural network for detection to verify the location of the hole, the size of the hole area, and the time when the hole appears.

[0015] (3) Compare the detection results of the heat transfer model data established by the finite element method with the detection results of the thermal imager experiment data to check whether the hole position error, area error, and occurrence time difference are less than e1, e2, and e3, respectively. If the result is "yes", the convective heat dissipation coefficient and heating power of the on-site environment are obtained. If the result is "no", the environmental coefficient and heating power are updated through Monto carlo, the finite element heat transfer model is re-established, and the experiment is carried out in sequence.

[0016] (4) The hole location, hole area size, and hole appearance time are obtained by using a convolutional neural network, and the convective heat dissipation coefficient and heating power of the on-site environment are obtained by comparing the finite element heat transfer model experiment and thermal images. The hole size, location, and hole depth are finally obtained by using an SVM support vector machine to extract features for training and detection.

[0017] (5) This method can obtain parameters such as temperature and power in the real experimental environment through reverse design, avoiding experimental changes and inaccuracies caused by slight changes in the environment.

Claims

1. A method for detecting thermal wave holes in varying environments based on a dual-drive approach of physics and data, characterized in that, The method combines a physical model and a digital model to achieve the final detection. The physical model consists of finite element simulation, a thermal imager, and the Internet. The digital model consists of a convolutional neural network and a support vector machine. The physical model is used to obtain raw data, which is then transmitted to the digital model via the Internet for training and detection. This ultimately achieves the detection of thermal wave holes in changing environments through a dual-drive approach of physical and data analysis. Includes the following steps: (1) Use the finite element simulation program to set environmental parameters, establish a finite element heat transfer model, then extract the simulation model image and transmit the data over the Internet; in the actual experiment, measure the size, location and depth of the hole; use a thermal imager to collect thermal images of the site, and then transmit the data over the Internet; transmit the data generated by the heat transfer model established by the finite element and the thermal images of the site collected by the thermal imager to the convolutional neural network for training and detection, and obtain the hole location, hole area size and hole appearance time; (2) Compare the detection results of the heat transfer model data established by the finite element method with the detection results of the thermal imager experiment data to check whether the hole position error, area error and occurrence time difference are less than e1, e2 and e3 respectively. If the result is "yes", the convective heat dissipation coefficient and heating power of the on-site environment are obtained; if the result is "no", the environmental coefficient and heating power are updated by Montocarlo, the finite element heat transfer model is re-established, and the experiment is carried out in sequence. (3) The hole location, hole area size, hole appearance time, and on-site convective heat dissipation coefficient and heating power obtained by using convolutional neural network and comparing the finite element heat transfer model experiment and thermal image are used to extract features for training and prediction, and finally the hole size, location and hole depth information are obtained.

Images