Cyber-Physical Systems (CPSs) face growing complexity, especially in safety-critical areas. Ensuring their correctness is vital to maintain full operational capacity, as undetected failures can be both costly and life-threatening. Therefore, advanced fault diagnosis procedures are essential for thorough CPS testing, enabling accurate fault detection, explanation, and rectification. This doctoral research contributes to the field by developing novel tools and techniques to enhance fault-based testing and diagnosis of CPSs. Our research focuses on testing of CPS dataflow models created in Simulink, validated against strict formal specifications. Our contributions include (i) an automated tool for systematic fault injection, (ii) a bio-inspired global optimization algorithm, (iii) a robust fault localization method, (iv) a novel approach to mutation testing for evaluating test suites against formal properties, and (v) a new coverage criterion tailored for CPS dataflow models. This comprehensive approach offers significant improvements over existing methods, ensuring thorough testing across various scenarios. We validate the effectiveness of our solutions using publicly available benchmarks from various domains. Our findings open new perspectives on CPS testing, laying the foundation for more robust CPSs.