We propose an Extreme Learning Machine (ELM) optimized by a quantum genetic algorithm formulated on Bloch-sphere Coordinates with Chaotic Mutation (CBQGA), hereafter referred to as CBQGA-ELM. CBQGA selects the ELM hyperparameters—including the number of hidden neurons, input weights, and hidden biases—by minimizing the Root Mean Square Error (RMSE) on a validation set. The resulting model is utilized to rolling bearing fault diagnosis. First, CBQGA-ELM is benchmarked against Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Quantum Genetic Algorithm (QGA), and Bloch-quantum Genetic Algorithm (BQGA) optimizer using datasets from the UCI Machine Learning Repository. Simulation results show that CBQGA delivers superior optimization performance relative to these alternatives. Second, laboratory experiments are conducted in which vibration signals are collected for four bearing conditions: healthy, inner race fault, outer race fault, and ball fault. Time-domain features are extracted from the signals and supplied to the diagnostic models. The results demonstrate that CBQGA-ELM achieves higher diagnostic accuracy and reliability than the competing methods, indicating its suitability for rolling bearing defect diagnosis.
