Agentic Systems for Failure Attribution and Root Cause Analysis

Sammanfattning

Automotive warranty cases are valuable for quality follow-up, but they are notwritten as root-cause reports. A record can mix customer symptoms, workshopactions, replaced assemblies, parts context, and partial causal clues. The recordedrepair scope can describe what was done in service while still hiding the componentthat most likely initiated the failure. Human reviewers can recover that meaning,but the work is slow and depends on engineering experience.This thesis aims to develop a planner-guided agentic system for failure attributionand root cause analysis in warranty cases. Instead of treating repair text as asingle classification input, the system separates text normalization, signal extraction,hypothesis formation, optional critique, dictionary-constrained mapping, historicalcaseretrieval, and final judgement. The intermediate outputs are stored with thefinal decision, so a reviewer can see the fields, comparison cases, and constraintsbehind the selected label.The evaluation combines manual review analysis, a direct single-pass baseline, mechanismvariants, field-level checks, retrieval checks, knowledge-use analysis, and finallabelscoring. Manual review finds 333 CHG-related records inside a 628-recordengine-exchange candidate set, showing how component-level failure patterns cansit inside a broad repair outcome. On the 357-record positive benchmark, targetlabelrecovery increases from 29.4% with the direct single-pass baseline to 84.3%with the planner-guided workflow, while other non-target outputs fall from 67.2%to 7.6%. On the mixed 628-record basis, the workflow reaches 85.4% accuracy and87.3% F1 for the target label. ANN retrieval preserves the exact top result in 99.4%of the current retrieval index, and knowledge entries are used in 490 of 628 completedrecords.The results indicate that an agentic workflow can improve failed-component recoveryand make the result easier to inspect than a single-pass request. It does not replaceengineering judgement or independently prove physical root cause. Its role is toorganize noise from workshops into a standardized label, structured support, and ashorter path for expert review.