Enhanced health states recognition for electric rudder system using optimized twin support vector machine

Safety and reliability represent indispensable prerequisites for electric rudder systems (ERS), while health states recognition serves as a potent technology that fortifies and optimizes these essential aspects. To address this problem, we present a health‐state recognition muti‐class model BAFAO‐IPBT‐TWSVM for ERS considering several typical operating parameters obtained from intelligent electric rudder system test platform. The twin support vector machine (TWSVM) not only possesses the ability of traditional fault diagnosis methods based on SVM to handle unbalanced data, but also introduces two non‐parallel hyperplanes to replace single hyperplane of traditional SVM. Traditional TWSVM simplifies and streamlines the problem‐solving, but it is limited to binary classification problem. Therefore, the improved separability principle weighting intra‐class distance and inter‐class distance generates the best decision tree structure named improved partial binary tree (IPBT) is to effectively decompose multi‐classification problem into multiple binary classification problems. A novel intelligent algorithms called bat algorithm‐based fruit fly optimization algorithm (BAFOA) is utilized to self‐adaptively optimize the parameters of each sub‐classifier TWSVMi. This strategic integration makes the model more flexible in adapting to the characteristics of electric rudder system and enhances the accuracy and robustness of the model. The performance of the proposed model is validated under real‐world datasets by the results of health states recognition experiments. The Accuracy, Precision, TPR, TNR, F1‐score, G‐mean, and Kappa of the BAFOA‐IPBT‐TWSVM are 0.972, 0.987, 0.982, 0.959, 0.985, 0.970, and 0.954 respectively. The reserved BAFOA‐IPBT‐TWSVM is capable of recognizing the health status with preferable performance compared with other nine models, which could introduce a novel idea for future rudder maintenance approaches.