An explainable hybrid deep learning-enabled intelligent fault detection and diagnosis approach for automotive software systems validation

Explainability is critical for safety-critical systems: The paper demonstrates that "black-box" AI models used to detect faults in automotive software lack interpretability, which "prevents the model from being adapted based on the prediction result" and "limits confidence in real-time safety-critical applications"—highlighting why understanding why an AI makes a decision matters as much as the decision itself.

Hybrid AI approaches can improve both accuracy and transparency: The researchers combined two types of neural networks (1D CNN and GRU) with explainability techniques (SHAP, DeepLIFT, and Integrated Gradients) to create a system that not only detects faults but also explains the reasoning behind each detection, enabling engineers to perform root cause analysis.

Explainable AI reduces computational costs and development burden: By making the fault detection model interpretable, the approach reduces "the computational cost required for developing a complex FDD model," suggesting that transparency can actually make AI systems more efficient and practical to deploy in real-world automotive validation.

Real-world testing validates responsible AI principles: The model was tested on actual data from hardware-in-the-loop simulations rather than synthetic datasets, demonstrating that explainable AI methods can work in genuine safety-critical automotive contexts where stakes are high.