As piston pin worn features are susceptible to environmental vibration disturbance during diesel engine operation, an effective vibration signal decomposition and noise reduction process is a promising way to enhance the disturbed signals, which is essential to build a reliable and precise binary classifier model to identify piston pin worn. To solve the problem of vibration signal decomposition and noise reduction, a feature extraction algorithm based on orthogonal empirical mode decomposition（ OEMD ）combined with continuous wavelet transform（ CWT ）and principal component analysis（PCA）was proposed. The orthogonal sensor layout was used to collect the vibration signal of the piston pin of the diesel engine in actual operation, and OEMD was used to decompose the orthogonal fusion vibration signal into multiple intrinsic mode functions（IMF）, and then the first four IMF components with 85% energy were selected for CWT processing to obtain the wavelet coefficient matrix. Finally, the optimal score matrix after PCA operation was input into the K-means clustering algorithm for classification. The actual experimental data verified the effectiveness of the proposed method, and the orthogonal fusion results integrated the overall trend and extreme value distribution, so it was more reliable than a single sensor, thus avoiding the interference or feature loss caused by inappropriate sensor installation position. Compared with EMD combined with AR spectrum algorithm and VMD algorithm, the proposed method had stronger noise reduction and feature extraction capabilities, and the classification effect was more obvious in K-means algorithm, which laid a foundation for two-classifier modeling and identification of piston pin wear.