This article develops a hybrid fault diagnosis (FD) method for industrial machines. The method considers the traits of reliable information from models and multidomain features of signals, and optimally incorporates the capabilities of intelligent processing techniques. The approach is inspired by a multilayer biological immune system, and consists of generalized (nonspecific) and specialized FD subsystems. First, a genetic algorithm-optimized artificial immune system technique is presented, which uses signal processing to extract multiperspective system features and selects low-dimensional features for intelligent fault detection. Second, a system identification approach is employed, which incorporates adaptive thresholding-based fault detection, and a fault severity index for fault identification. The developed hybrid FD technique opts for a synergy-based coordination approach of nonspecific intelligent fault detection and specific model-based FD. Specifically, it analyzes the data-parallel operation of the two methods and incorporates a comprehensive self-assessment-based conflict resolution mechanism to achieve improved and reliable FD in case of incomplete dataset knowledge and model discrepancies. The efficacy of the developed method, in FD, is assessed using systems with broken rotor and bearing fault.