Diagnosis of Diesel Engine Based on Oil Analysis Using Convolutional Neural Network and Visualization

Oil condition monitoring is an effective method for detecting abnormal erosions or defects in mechanical equipment and systems. One of the issues in the field of condition monitoring with the help of oil analysis is the cost and time required to inspect all samples by an expert. However, not all oil analysis samples need to be reviewed by an expert, and less than 10% of this data indicates a critical situation that requires rapid planning and action. The goal of this article is to turn the oil status into an image so that you can quickly identify the oil status by looking at the image. Also, by processing these images by software, the state of failure can be extracted through artificial intelligence. In this research, the data are taken from the sample test of motor oils of road construction rollers. First, the data were converted to gray scale images using baseline for diesel engines and through MATLAB software. In the next step, these images are processed using the convolution neural network method to determine the oil status. Comparison of the obtained results showed that the visualization of the oil analysis results helps to understand the general condition of the oil for the user and the critical samples and the need for action are identified more quickly among the mass of oil samples.