High-precision visual inspection of semiconductor optical devices is critical but challenged by the coexistence of local and global anomalies, alongside domain-specific imaging artifacts. This study proposes a frequency-specialized dual-branch anomaly detection framework integrated with domain-specific preprocessing to address these issues. First, a Masked Autoencoder (MAE)-based preprocessing module reconstructs peripheral shadow artifacts to eliminate boundary-induced false positives. Then, utilizing the Stationary Wavelet Transform (SWT), the method explicitly decouples defect signatures into frequency domains. Low-frequency components are routed to an MAE-based global branch, while high-frequency components are processed by a Reverse Distillation (RD++)-based local branch. Experimental results on an industrial dataset demonstrate that the proposed method achieves an AUROC of 0.9837 and an AP of 0.9701, significantly outperforming the baselines. By attaining a 100% detection rate for well-defined local defects, the framework demonstrates its high reliability for strict industrial quality control where minimizing false negatives is paramount.

This study investigates proactive scheduling for production spike mitigation in semiconductor manufacturing under temporary capacity fluctuation caused by production machine failures and recoveries. A phase-dependent control framework is developed with two mechanisms: machine allocation control and release control. Machine allocation control switches policies across production phases to reduce queue-time constraint (Q-time) violations and inter-product imbalance, while release control adjusts release quantities during predicted degradation periods to limit work-in-process (WIP) accumulation and waiting-time violations. Degradation periods are identified using either a data-driven failure-probability model based on machine data or a Weibull-based probabilistic model. Performance is evaluated in terms of throughput, cycle time, waiting time, variability, and rework rate. Both proactive strategies outperform the reactive baseline. Moreover, the data-driven approach reduces cycle time, waiting time and rework while maintaining stable throughput. These results demonstrate that degradation-aware proactive scheduling can mitigate production spikes and improve operational performance in non-stationary semiconductor manufacturing environments

This study presents an automated tool validation framework developed within Siemens NX CAM using NX Open Python scripting to improve tooling consistency and workflow standardization in CNC programming environments. The framework integrates sequential tool renumbering, duplicate tool detection, diameter validation, material classification, and unused-tool cleanup through rule-based automation logic. The implementation was evaluated using multiple CAM files with varying tooling complexity through before-and-after validation analysis. Initial results demonstrated improvements in tooling-data organization, reduction of duplicate-tool conditions, and enhanced workflow consistency. The proposed framework supports digital manufacturing and intelligent manufacturing initiatives by reducing repetitive manual verification tasks and improving tooling-data reliability within CAD/CAM workflows.

Recent advances in large language models (LLMs) have created new opportunities for automating computeraided design (CAD) workflows through natural language interactions. However, generating reliable CAD scripts remains challenging due to execution errors, geometric inconsistencies, and increasing design complexity. This paper presents an iterative AI-assisted CAD generation workflow that combines prompt engineering, CAD validation, and corrective feedback to improve the reliability of AI-generated CAD scripts. The proposed workflow evaluates ChatGPT-5, Claude, and Copilot across three CAD geometries of increasing complexity using both CQ-Editor and FreeCAD environments. Generated scripts are executed and validated for geometric correctness, watertightness, STL compatibility, and manufacturability. When execution or geometry errors are detected, visual and textual feedback is used to iteratively refine the generated scripts through a structured correction process. Experimental results show that iterative feedback significantly improves CAD generation success rates across all evaluated models. Claude was the most consistent across CAD environments and shapes; ChatGPT generally needed fewer corrections for complex shapes; and Copilot showed greater difficulty, especially when it came to more complex shapes. These results highlight the strengths and weaknesses of AI-assisted CAD generations and demonstrate how iterative feedback improves CAD script reliability.