This chapter addresses the issue of sample value imbalance in process monitoring by developing a fault detection approach that integrates variable selection and support vector data description (SVDD) for enhanced efficiency. Initially, KLD is employed as the variable selection algorithm to identify the most informative variables related to specific faults. These selected variables are then segmented into blocks to prevent fault information from being obscured within a single monitoring space, ensuring that relevant variables and critical information are concentrated within the same block. Subsequently, kernel principal component analysis is conducted within each block to manage the challenges posed by high-dimensional and nonlinear variables. The monitoring outcomes are derived using the proposed SVDD, which incorporates a restructured radius index that exhibits heightened sensitivity to faults. Experimental results on the Tennessee Eastman process demonstrate the efficacy of this method, highlighting its superior mean fault detection rate compared to existing approaches.