Objectives

  • This is my final year capstone project that I did with an industry partner - Kelco - An industrial pump controller manufacturer base in Sydney (Brookvale).
  • Kelco needs me to create a system that help them automatically calibrate their pressure sensor after the assembly process, and testing the relays under 220V input.
  • Before, it require a person to do it because the pressure sensor can be calibrated by pumping a pressure of 500kPA and inspect the real value on the LCD screen of the pump controller. After that, he will decide to press the up or down arrow button to adjust the inspected value to match with the true value.
  • Secondly, after being calibrated, it will come to the high-voltage testing to see if the relay work fine with the 220V input voltage. If it work fine, there are a pattern of 3 LEDs blinking on the pump controller. The inspector must inpsect the pattern and give a decision.
  • My approach for this problem is using computer vision especially OCR technique (Optical Character Recognition) to read the text on the LCD screen to calibrate the pressure sensor. After that, I will use image processing technique to record the pattern of LED blinking and compare with the correct pattern.
  • Overall, my role is a combination of mechanical design for the system, electronics design for the control system, firmware design for the system, software UI design for the HMI interface.

    • Physical system

  • This system is still in developing stage, so the wiring and electric box are still a bit messy.
  • Overall, the system still can fulfill the requirements where safety is the first priorirty.

    • 3D Model Design

  • In this design, I have ultilized the DFMA and DFAM to design and manufacture all the parts in this project. I have used 2020 aluminum profile to build the main frame and I have printed most of the other part with PLA and ABS for the parts that under more stress.
  • The advantage of this design is that, it is easy to disassemble and assemble again but still keep a good precision. Most of the bolts are M5 profile, so that it is easier to maintain.

    • Dev GUI Interface

  • This is the dev version of the GUI used to test the OCR model. I have use Qt Creator and C++ to develop the GUI. It communicates with the ESP32 via UART.

    • Demo video

    PCB Design and Firmware Production

  • I also take responsible for creating the firmware for the pump controller. In this one, they use the PIC18F67K22 of Microchip. Before, the firmware was made by the PIC block programming, it consumes a lot of memory which is 80%. When I create the firmware with MPLab IDE in C language, it reduce to 40% of total memory. Overall, re-creating the firmware in a universal language such as C helps to reduce the cost for the MCU, and making the firmware maintainance and update easier for the future engineer.