Random Convolutional Kernels Improve Machinery Fault Diagnosis

Ensuring the reliability of industrial processes heavily depends on accurately diagnosing faults in rotating machinery. Random convolutional kernel-based Time Series Classification (TSC) methods, such as ROCKET and its variants, have gained traction due to their favorable balance between predictive performance and computational efficiency. This study focuses on evaluating SelF-Rocket for the multi-class diagnosis of both mechanical and electrical faults.A key contribution of this work is the introduction of a multivariate extension to the original SelF-Rocket method. The enhanced approach was rigorously compared against other leading ROCKET-based methods using two public benchmark datasets: MaFaulDa, which covers mechanical faults, and ITSC-UDG, which addresses stator inter-turn short circuits. These comparisons were conducted under both univariate and multivariate settings to provide a comprehensive assessment.The experimental results clearly demonstrate that SelF-Rocket achieves the best overall accuracy-latency trade-off among the evaluated methods. It delivered the highest classification performance on the MaFaulDa dataset and remained highly competitive on the more challenging ITSC-UDG dataset. This indicates that SelF-Rocket offers a robust and efficient solution for fault diagnosis in industrial machinery.