Distributed Hyperdimensional Computing for Real-Time Data Aggregation and Interpretable Quality Monitoring in Manufacturing
IMECE 2024 (Porland, OR) [Under Review]
Zhiling Chen, Danny Hoang, Ruimin Chen, Farhad Imani.

Abstract: The integration of diverse sensors in manufacturing processes offers enhanced potential for real-time quality assurance through the collection of complementary data. However, the limited interpretability of current machine learning models often hampers the effective discrimination of each sensor's unique contributions, primarily due to the complexity of decoding the interdependencies among various signals. This paper proposes a novel computational framework, Distributed Hyperdimensional Computing (DHDC), which is designed to leverage efficient cognitive operations, such as binding and bundling, for interpretable learning across multi-level sensor data. DHDC operates by encoding and aggregating data systematically across a distributed architecture, thereby enhancing transparency in computational efficiency. Our real-world experimental results on the 5-axis machining for the fabrication of the counterbore hole feature demonstrate that the framework not only effectively characterizes the impact of individual sensors but also achieves a high degree of predictive accuracy, as evidenced by an F1 score of $90.4 \pm 0.013\%$. The proposed framework holds the potential for interpretable and scalable quality control in distributed additive and subtractive manufacturing. ... See More