Fault Detection and Explainable Classification in Automotive HIL Validation via Denoising Autoencoders and In-Context Large Language Models

2026-07-04

AI Summary

This work addresses the limitations of traditional hardware-in-the-loop (HIL) validation for automotive systems, which relies on handcrafted rules and supervised learning, thereby struggling to detect unknown faults and being constrained by labeled data and poor interpretability. The authors propose a two-stage unsupervised anomaly detection and few-shot classification framework. First, a denoising autoencoder trained exclusively on normal signals enables high-generalization anomaly detection via reconstruction error, achieving F1 scores of 0.97–0.98 and mean errors below 0.03. Subsequently, anomalous signal windows are transformed into textual statistical evidence and fed into a frozen large language model (e.g., Mistral Small 24B) for few-shot prompt-based classification, yielding fault type, confidence, location, and natural-language explanations. This approach eliminates dependence on fault labels and predefined rules, striking a favorable balance among accuracy, interpretability, and inference efficiency.

Abstract

Validating automotive software systems produces large multivariate test recordings that are still examined through effort-intensive manual review and rule-based evaluation, which detects faults beyond predefined rules poorly. Machine and deep learning have advanced fault diagnosis, yet most supervised models require large labelled datasets, generalise poorly to unseen conditions, and offer little insight into their decisions. We propose a generalisable and explainable two-phase framework for fault detection and classification during real-time validation. A denoising autoencoder trained only on healthy signals first flags abnormal behaviour through reconstruction-error analysis, removing the need for fault labels. Each abnormal window is then encoded as compact textual statistical evidence relative to a time-aligned healthy reference and classified by a frozen large language model under zero-shot and few-shot prompting, returning the predicted class, ranked alternatives, a confidence value, the fault location, and a short evidence-based explanation. Eight open-source models are evaluated across two powertrains and three driving regimes. The detector attains average F1-scores of 0.97 across powertrains and 0.98 across regimes, with average mean error below 0.03. Zero-shot prompting proves insufficient (best 0.519 F1-score), whereas few-shot prompting reaches perfect discrimination under stable regimes, showing that prompting strategy, rather than parameter count, governs classification quality: a nine-billion-parameter model surpasses every zero-shot medium and large model. Mistral Small 24B is adopted as the main pipeline model for its balance of accuracy, class-balanced reliability, calibration, and inference cost, giving engineers interpretable diagnostic reports and more efficient validation.

Problem

• fault detection
• explainable classification
• automotive HIL validation
• generalisation
• interpretability

Innovation

• Denoising Autoencoder
• Large Language Model
• Explainable AI
• Few-shot Prompting
• Fault Detection