Classification is an important technique used for condition monitoring. Extensive research has been carried out on classification and numerous techniques have been developed. The support vector machine (SVM) is one of these techniques; it has excellent classification capacity and is widely used. The effectiveness of the SVM depends on the selection of the kernel function, so to maximise performance this paper proposes using an adaptive multiple kernel SVM (AMK-SVM). Using AMK, many heterogeneous features, such as continuous, categorical, logical etc, can be merged. Instead of predefining the parameters of kernel functions as with other multiple kernel SVMs, this method can adapt its parameters to data automatically through kernel alignment. The paper offers two numerical examples: one with benchmarking data to test the feasibility and performance of the approach (for this case a two-layer neural network and two single kernel SVMs are applied to the same datasets to compare their performance with the AMK-SVM); the other example uses the AMK-SVM to discriminate a healthy bearing from a defective bearing