Scholarlink

<p>Achieving zero-defect manufacturing (ZDM) in semiconductor assembly requires root-cause analysis (RCA) methods that do more than flag excursions. They must produce explanations and corrective actions that are traceable to process evidence, interpretable to quality engineers, and maintainable within governed manufacturing workflows. This paper presents a deterministic, provenance-aware RCA framework for surface-mount technology (SMT) assembly that transforms heterogeneous line data, including specifications, measurements, inspection outcomes, and troubleshooting knowledge, into explicit causal-chain representations connecting parameter deviations to mechanisms, defects, and corrective-action specifications. The framework operationalizes three recurring reasoning patterns: Specification–Observation–Conformance Assessment, Disposition–Realization–Failure Cause, and Cause–Effect–Corrective Action. These patterns are implemented over a governed knowledge graph using SPARQL rule templates that materialize conformance assessments, failure causes, directed causal links, and provenance traces. We demonstrate the approach on solder-bridging and open-circuit scenarios using synthetic data designed to reflect industrial SMT data schemas and industrially plausible value regimes. System inferences are compared against expert-reviewed reference graphs using chain precision and recall, provenance completeness, and cycle rate. Across 30 simulated shifts of 5000 printed circuit boards (PCBs) each, the framework achieves approximately 90% chain precision and recall with 100% provenance completeness and no detected cycles. Under combined measurement noise and missingness, chain recall and precision decrease to approximately 81%, indicating reduced but still reviewable diagnostic fidelity under degraded observations. These results demonstrate that deterministic, human-interpretable RCA chains can be generated and retained as auditable engineering knowledge under controlled, industrially informed conditions. The study is positioned as a semantic and methodological foundation for governed RCA decision support, not as a production-qualified deployment; retrospective validation on historical SMT defect logs and expert-in-the-loop pilot assessment remain necessary next steps.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An ontology-based approach to root-cause analysis for zero-defect surface-mount assembly

  • 期刊论文

  • 发表日期：

    2026年7月22日

查看全文
  • Mukund Shenoy,
  • Farhad Ameri

摘要

Achieving zero-defect manufacturing (ZDM) in semiconductor assembly requires root-cause analysis (RCA) methods that do more than flag excursions. They must produce explanations and corrective actions that are traceable to process evidence, interpretable to quality engineers, and maintainable within governed manufacturing workflows. This paper presents a deterministic, provenance-aware RCA framework for surface-mount technology (SMT) assembly that transforms heterogeneous line data, including specifications, measurements, inspection outcomes, and troubleshooting knowledge, into explicit causal-chain representations connecting parameter deviations to mechanisms, defects, and corrective-action specifications. The framework operationalizes three recurring reasoning patterns: Specification–Observation–Conformance Assessment, Disposition–Realization–Failure Cause, and Cause–Effect–Corrective Action. These patterns are implemented over a governed knowledge graph using SPARQL rule templates that materialize conformance assessments, failure causes, directed causal links, and provenance traces. We demonstrate the approach on solder-bridging and open-circuit scenarios using synthetic data designed to reflect industrial SMT data schemas and industrially plausible value regimes. System inferences are compared against expert-reviewed reference graphs using chain precision and recall, provenance completeness, and cycle rate. Across 30 simulated shifts of 5000 printed circuit boards (PCBs) each, the framework achieves approximately 90% chain precision and recall with 100% provenance completeness and no detected cycles. Under combined measurement noise and missingness, chain recall and precision decrease to approximately 81%, indicating reduced but still reviewable diagnostic fidelity under degraded observations. These results demonstrate that deterministic, human-interpretable RCA chains can be generated and retained as auditable engineering knowledge under controlled, industrially informed conditions. The study is positioned as a semantic and methodological foundation for governed RCA decision support, not as a production-qualified deployment; retrospective validation on historical SMT defect logs and expert-in-the-loop pilot assessment remain necessary next steps.