Monitoring Valve and Spark Plug Failures using Z-Freq Statistical Analysis

Abstract

Engine problems such as spark plug misfire, valve clearance and valve cracks are the failures that lead to engine malfunction if engine continues in operation and require early detection. Machine learning can be used to automatically diagnose engine problems, but a high success rate is still a barrier because a significant amount of training data is needed. Therefore, in order to examine the specific frequency characteristics, this new fault analysis approach will take into frequency content throughout the Fast Fourier Transform (FTT) procedure. Z-freq, a dynamic vibration signal analysis for non-deterministic data that focuses on frequency domain rather than time data, is a new statistical signal-based analysis that can easily be used to improve these vibration-induced faults studies at this stage for further enhanced investigation. This examination examines the time and frequency domain of data acquired from Proton, Toyota, and BMW engines operating at various speeds between 750 and 3000 rpm. In order to replicate the real effect, the fracture fault was made using a wire cut method for 0.25, 0.5, and 1.0 mm, while the clearance fault was set using a screw and sheet gauge for clearance thicknesses of 0.0, 0.2, 0.3 mm during experimental tasks. The voltage applied to a particular spark plug will be cut to simulate a misfire issue. High sensitivity, space-saving, and long-lasting piezo-based sensors are used to assess vibration caused by spark plug misfire, valve clearance and valve crack. Piezo-film, micro-fiber composite, and accelerometer sensors are employed in this experiment. To guarantee precise and accurate observation, all of these sensors were calibrated using the Bruel and Kjaer type 4294 calibration exciter. The main finding is that the distribution of Z-freq data for normal, misfire, valve clearance, and valve crack shows a notable pattern in the coefficient value.