The thesis addresses development of a test bench that would enable the discovery of rotor flaws. Since the rotors were not to be destroyed during the testing, non-destructive testing methods were opted for. Two key methodologies used during the development of the test bench were resonance frequency analysis and machine learning.
Multiple types of rotors and magnets were tested with the use of acoustic emission. The test samples were stimulated and their response was received with a microphone. A program which calculates the frequency spectrum of the measured units with the help of Fourier transform and estimates their quality from the position of resonant frequencies was developed in the Labview. The test bench was upgraded with the use of machine learning, which was realised using the Weka software.
The results show that the acoustic emission method is appropriate for the testing fuel pump rotors. Multiple types of rotors and magnets were tested, discovering that the chosen method is not appropriate for all types of test samples. The reasons for this result are unfavourable shape of the measured units and too indistinctive flaws.The work was concluded by practically carrying out quality control in the production process.
|
