Academic Publication engineering thesis cnc electron-beam welding sapientia opencv

Fissure recognition and CNC controlled electron-beam welding

Author: Kodok Márton
Institution: Sapientia – Hungarian University of Transylvania, Târgu Mureș, Romania
Type: Final Year Project / Thesis (2008)
Original Publication Date: April 19, 2010

Abstract

In this paper, the author used intensity pictures for recognizing fissure trajectories. The pictures were recorded by an electron-beam machine, projected and developed by PhD. Dávid László.

These images have been enhanced with complex filters and image recognition procedures used in the digital processing industry to extract fissure trajectories in a thin and continuous format. Based on the digital information from the filtered images, another program, called postprocessor, generated a CNC numerical algorithm for following the fissure trajectory.

One deficiency of the electron-beam machinery is the missing precise and modern positioning system. Based on the industry requirements of electron-beam processing under vacuum, an XY linear positioning stage was chosen to hold the metal and move it under the beam. Graphical user interface software has been created to transcribe CNC numerical codes into the industrial language understood by the motion controllers. Therefore, the piece-holding table can be controlled using CNC algorithm codes.

Electron-beam applications are used more and more in the automotive manufacturing industry. The exiting electron jet from the electron gun is focused on the metal piece with electromagnetic lenses. Because of its great focusing feature and the amount of energy density that can be achieved in a small spot, it leads to a precise manufacturing process compared to universal methods. Electron-beam installations are equipped with image recording equipment; the retro-diffused electrons from the surface are used for analog processing and converted to digital data, creating a digital picture on the computer.

Nowadays in our digital and automatic way of life, using a computer to extract information from digital sources for industrial processes is common and modern. From the images taken with the electron-beam machine, image processing techniques extract the fissure trajectory. OpenCV is an open-source computer vision library focused mainly on real-time image processing. A few years ago, programmers had to write thousands of lines of code to get the desired results; with the OpenCV library, this is reduced significantly, making programmers' lives easier and boosting image processing usage in industry. From the machine-taken intensity pictures, noise is removed with well-chosen filters and structural elements. The filtered image is modified with morphological image pattern recognition instructions, resulting in a thin and continuous trajectory of the fissure.

For practical usage, a graphical user interface desktop application was created. In this application, users can fine-tune the parameters used in image processing. On a metal piece, fissures can occur near important edges. In this situation, the edge path can be seen alongside the fissure path on the picture. To achieve good results, the user can define an area on the image to be omitted. Recognized fissure data in vector format is used for the CNC postprocessor part of the desktop software, generating CNC numerical codes to follow the fissure's trajectory.

One deficiency of the electron-beam machinery is the missing precise and modern positioning system. Based on CNC machine tools, upgrading to a positioning system in a 2-axis configuration moving the metal under the electron beam is a major improvement. This stage is controlled numerically with CNC codes.

The e-gun welding process takes place inside a closed vacuum chamber. The presence of vacuum justifies constructing the stage from special materials, taking into account outgassing effects, demagnetization, and degreasing. In the chamber, high temperatures, splashes of material micro-particles, and X-rays (Roentgen rays) must be considered when choosing operating devices. Stage selection must also account for geometrical and technological factors, such as using vacuum-compatible lubricants and ensuring optimal motor performance.

The manufacturer provides motion controllers for the 2-axis XY stage. The motion controller can be programmed from a personal computer via the RS-232 serial standard. The created GUI desktop application generates CNC codes based on the extracted trajectory of fissures and transcribes CNC codes into the industrial language understood by motion controllers.

The results of the presented paper include fissure recognition, generation of CNC codes to follow the path, the selection process of a stage for specific environmental conditions, and the ability to numerically control this stage with CNC codes. Future work considers implementing the whole system as a CNC machine tool and enhancing the image recognition algorithm to extract more complex forms.

Table of Contents

  • Romanian Table of Contents (page 1)
  • Hungarian Table of Contents (page 4)
  • English Table of Contents (page 7)
  • Romanian Abstract (page 10)
  • Hungarian Abstract (page 12)
  • English Abstract (page 14)

1. Introduction (page 16)

2. Presentation and Usage of the Electron-Beam Machine (page 17)

  • 2.1. Importance of electron-beam applications in manufacturing industry (page 17)
  • 2.2. Theory of electron-beam process (page 19)
    • 2.2.1. Acceleration of electrons (page 19)
    • 2.2.2. Surface heating and particles vapors in e-gun process (page 20)
  • 2.3. Structure of electron gun machines (page 22)
  • 2.4. The used electron-beam machine (page 25)
  • 2.5. Electron-gun (page 27)
  • 2.6. Vacuum creating system (page 29)
  • 2.7. Image converters (page 31)
  • 2.8. Dimension of images on the piece surface (page 32)

3. Processing of the Images with Custom Algorithm (page 34)

  • 3.1. Image processing with computers (page 34)
  • 3.2. Digital images in computers (page 35)
  • 3.3. Image recognition theory bases (page 36)
  • 3.4. Arithmetical operations with digital images (page 37)
  • 3.5. Thresholding (page 38)
  • 3.6. Translation (page 39)
  • 3.7. Vector type operations – average intensity (page 39)
  • 3.8. Convolution filters (page 40)
    • 3.8.1. Low pass filter (page 41)
    • 3.8.2. High pass filter (page 41)
  • 3.9. Morphological operations (page 41)
    • 3.9.1. Dilation (page 42)
    • 3.9.2. Erosion (page 43)
    • 3.9.3. Opening and closing (page 43)
    • 3.9.4. Objects skeleton (page 44)
  • 3.10. Structure of the own image processing program (page 46)
    • 3.10.1. The aim of the image data and program input (page 48)
    • 3.10.2. Idea of preprocessing (page 50)
    • 3.10.3. Tried analyses on the filtered image (page 54)
    • 3.10.4. Dilation of the filtered image (page 56)
    • 3.10.5. Searching of the skeleton (page 57)
    • 3.10.6. Postprocessing - conclusions (page 60)

4. Metal Holding XY Stage Selection Process and CNC Numerical Control Planning (page 61)

  • 4.1. The current position system of the welding machine (page 61)
  • 4.2. Upgrading of the position system (page 63)
  • 4.3. Review of numerically controlled machines (page 64)
  • 4.4. Selection of the 2 axes XY stage (page 65)
    • 4.4.1. Geometrical and process parameters used for selection (page 65)
    • 4.4.2. Ability to work in vacuum environment (page 66)
    • 4.4.3. Ability to work in high temperature environment (page 66)
    • 4.4.4. Demagnetization and degreasing factors (page 68)
  • 4.5. Technical sheet of the selected stage (page 69)
  • 4.6. Subparts of the linear stage (page 70)
    • 4.6.1. Linear guides (page 70)
    • 4.6.2. Ball screws (page 71)
    • 4.6.3. Encoders (page 71)
    • 4.6.4. Stepper motors (page 71)
  • 4.7. CAD model of the XY stage (page 72)
  • 4.8. Communication between stage and motion controller (page 74)
  • 4.9. Motion controller interpolation system (page 77)
  • 4.10. Placement of CNC control in the system (page 78)

5. Usage of Vectorized Images to Control Stage Motion (page 79)

  • 5.1. Presentation of the developed GUI application (page 79)
  • 5.2. Fissure trajectory transformation into vector list structure (page 80)
  • 5.3. Bases of CNC programming (page 80)
  • 5.4. Generating CNC numerical control codes to follow fissure path (page 82)
  • 5.5. Transcribing CNC codes into industrial language understood by motion controllers (page 83)

6. Results and Development Opportunities (page 85)

7. Bibliography (page 86)

  • Appendix (page 88)