The primary purpose of a hydraulic cylinder-type hydraulic system is to transform the hydraulic energy of the system into mechanical energy and enable the hydraulic cylinders linear reciprocating motion. The stability and effectiveness of the hydraulic system will be significantly impacted by hydraulic cylinder leaks, which are a common issue with these devices. In order to solve this issue, this paper suggests a weighted support vector machine with genetic algorithm optimized linear kernel and Gaussian kernel approach for diagnosing hydraulic cylinder leakage faults. Afterwards, this method is coupled with the One vs Rest approach to achieve multi-classification of hydraulic cylinder leakage faults. According to the experimental findings, the method has significantly improved fault classification accuracy compared to the single kernel SVM model and has good applicability in multi-classification tasks.