Abstract: Modern industrial manufacturing usually involves multimode processes but the traditional support vector data description (SVDD) algorithm is a single-mode fault detection method and is therefore unsuitable for multimodal processes. To address the above shortcoming, we propose a multimode process monitoring method that combines the nearest neighbors difference (NND) and SVDD (NNDSVDD) algorithms. First, we use NND method to eliminate the multimode structure of an original data set and to obtain a difference data set. Then, we implement the SVDD algorithm in the difference data set to determine the statistical control limit. Lastly, we calculate the statistical values of the test data and compare them with the control limit to determine the test data as being in either a normal or fault state. NND eliminates the multimodal data structure and provides a good data foundation for SVDD. By applying SVDD to difference data sets, we improve the fault detection ability of SVDD in multimodal processes. We applied the NNDSVDD in numerical simulation examples and a semiconductor manufacturing process to verify the performance of our proposed method. The fault detection rate of the NNDSVDD reached 100% in each case and the simulation results indicate that the proposed method outperforms the conventional SVDD.