Application of low-cost ADXL1002 accelerometer for vehicle engine misfire detection using a novel hybrid EMD-based image processing and DCNN-LSTM model

Engine misfires significantly impact vehicle performance and efficiency. This paper presents a novel misfire detection method using vibration data from an ADXL1002 accelerometer. The proposed approach employs Empirical Mode Decomposition (EMD) to effectively extract relevant frequency components from vibration signals, enhancing feature separation. The processed data is then transformed into 2D grayscale images and fed into a hybrid Deep Convolutional Neural Network–Long Short-Term Memory (DCNN-LSTM) model for classification. Experimental results showcase outstanding performances, achieving 100% training accuracy and 98.7% test accuracy. Comprehensive evaluation metrics—including sensitivity, specificity, balanced accuracy (BA), and geometric mean (GM)—validate the model’s robustness in detecting misfires across diverse operating conditions.