INTELLIGENT CAR DIAGNOSTICS USING ARTIFICIAL INTELLIGENCE

Abstract

This paper presents a novel approach that leverages On-Board Diagnostics (OBD-2) technology integrated with artificial intelligence (AI) to develop an advanced vehicle monitoring system capable of real-time diagnostics and predictive maintenance. With modern automobiles becoming increasingly complex, traditional diagnostic systems often fail to detect subtle deviations in engine performance until a critical failure occurs. Our study aims to bridge this gap by combining rich data extracted from OBD-2 interfaces with robust AI algorithms, thereby providing drivers and technicians with early warnings and actionable insights to prevent costly repairs and enhance vehicle longevity. The proposed system utilizes various key diagnostic parameters provided by OBD-2, such as ignition timing advance, knock sensor readings (both voltage and retard), as well as short-term fuel trim (STFT) and long-term fuel trim (LTFT). These parameters are critical for assessing engine performance, as they can reveal the quality of fuel used and potential issues like improper combustion or engine knock. By establishing baseline performance metrics under optimal operating conditions, our method employs statistical and machine learning techniques to detect deviations that may indicate underlying problems. Data is continuously collected from the vehicle’s electronic control units and preprocessed to eliminate noise and inconsistencies. Logistic regression is applied to classify the quality of fuel based on the observed sensor readings, while linear regression models predict adjustments in ignition timing. Furthermore, to capture the complex, nonlinear relationships inherent in the data, we implement a neural network model that improves the overall diagnostic accuracy. Experimental results demonstrate that our integrated system not only accurately classifies fuel quality but also reliably predicts potential engine malfunctions, offering early alerts and recommending corrective actions.