Intelligent prediction of engine failure through computational image analysis of wear particle

•Lubricating oil samples from CNG driven engine were collected and ferrography testing was conducted.•Image processing techniques are used for the morphological study of the wear particles.•A neural frame work model was designed to predict internal failure inside engine components.•The ANN based model with topology (3-8-5) gives an optimal results to predict failure. The present study is focused on studying the morphological characteristics of wear particles. The raw image captured from the CCD microscope is processed through a series of image processing techniques to obtain the final image. Fractal computation is performed to estimate the surface roughness of the particle boundary, which indicates the wear rate failure. An intelligence-based ANN model has been created using feed-forward backpropagation to predict outputs such as Form Factor, Convexity, Aspect Ratio, Solidity, and Roundness with respect to Running Hour, Engine RPM and Engine Oil temperature. The propound ANN model is seen as equipped to map the input-output patterns of failure under the engine parameter platform. Statistical analysis of combined error and correlation factor provides a powerful mapping tool for failure prediction. Subsequent ANN models have been seen as practical tools for predicting the performance of wear properties with the nominal inspection.