A unique technique is proposed using Principal Component Analysis (PCA) and softmax regression for automated fault detection and classification using vibration and acoustic signals generated from the IC engines. This technique uses the PCA for feature extraction and dimensionality reduction of the frequency spectrum data of noisy acoustic or vibration signals. These feature vectors are then used to get correlation coefficients of training, and testing data. These correlation coefficient vectors are used by soft max regression based classifier for classification of the engine into different classes. The proposed technique does not require any hand-engineered feature extraction, as usually done and no pre-filtering is required on noisy industrial data. The proposed technique is independently tested on two different types of data sets from the simulator and industrial environment. It has the performance of more than 98\% on vibration data from mechanical fault simulator for five different types of faults. The performance of this technique for acoustic data from industrial IC engine is more than 99\% for five different fault classes. In a typical case of industrial IC engines, for 216 test data sets, the classification performance is 99.54\% with only 72 training data sets.