This paper presents a case-based reasoning (CBR) approach to assist diagnosis problem based on the extracted features of engine ignition signals. The advantage of the CBR system is that multiple possible diagnoses can be provided to the user, instead of a single most probable diagnosis provided by the traditional network-based classifiers such as multi-layer perceptrons (MLP). To overcome the retrieval inefficiency of CBR, wavelet packet transform (WPT) and K-means clustering (KM) are integrated together to form a new signal preprocessing method for the CBR system. Experimental result shows that CBR using WPT and KM generates higher diagnostic accuracy than MLP.