NASPrecision: Neural Architecture Search-Driven Multi-Stage Learning for Surface Roughness Prediction in Ultra-Precision Machining

TL;DR

This work tackles the problem of predicting ultra-precision surface roughness under limited and imbalanced data. It introduces NASPrecision, a neural architecture search–driven, multi-stage learning framework that automates architecture discovery, initial prediction, and subsequent refinement, while employing generative data augmentation via a variational autoencoder and polynomial feature augmentation to improve robustness. Empirical results across three real-world machining datasets show substantial improvements in MAPE, RMSE, and STD ($18\%$, $31\%$, and $22\%$ on average, respectively) over strong baselines, with ablation confirming the benefit of the refinement stage. The proposed approach reduces reliance on domain expertise and enhances predictive accuracy, enabling more efficient production planning in optics, aerospace, and related industries.

Abstract

Accurate surface roughness prediction is critical for ensuring high product quality, especially in areas like manufacturing and aerospace, where the smallest imperfections can compromise performance or safety. However, this is challenging due to complex, non-linear interactions among variables, which is further exacerbated with limited and imbalanced datasets. Existing methods using traditional machine learning algorithms require extensive domain knowledge for feature engineering and substantial human intervention for model selection. To address these issues, we propose NASPrecision, a Neural Architecture Search (NAS)-Driven Multi-Stage Learning Framework. This innovative approach autonomously identifies the most suitable features and models for various surface roughness prediction tasks and significantly enhances the performance by multi-stage learning. Our framework operates in three stages: 1) architecture search stage, employing NAS to automatically identify the most effective model architecture; 2) initial training stage, where we train the neural network for initial predictions; 3) refinement stage, where a subsequent model is appended to refine and capture subtle variations overlooked by the initial training stage. In light of limited and imbalanced datasets, we adopt a generative data augmentation technique to balance and generate new data by learning the underlying data distribution. We conducted experiments on three distinct real-world datasets linked to different machining techniques. Results show improvements in Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Standard Deviation (STD) by 18%, 31%, and 22%, respectively. This establishes it as a robust and general solution for precise surface roughness prediction, potentially boosting production efficiency and product quality in key industries while minimizing domain expertise and human intervention.

Figures (9)

• Figure 1: The proposed NAS-driven multi-stage learning framework, NASPrecision
• Figure 2: Semantic meanings of the Search Space
• Figure 3: Abstract illustration of Neural Architecture Search methods elsken2019neural. A search strategy selects an architecture $\mathcal{A}$ from a predefined search space $S$. The architecture is passed to a performance estimation strategy, which returns the estimated performance of A to the search strategy.
• Figure 4: Different order of feature augmentation for MJP dataset
• Figure 5: Different order of feature augmentation for CNC Turning dataset