[Paper Review] A White-Box SVM Framework and its Swarm-Based Optimization for Supervision of Toothed Milling Cutter through Characterization of Spindle Vibrations

Tejas Y. Deo, Abhishek D. Patange+7 more|arXiv (Cornell University)|Dec 15, 2021

Advanced machining processes and optimization4 citations

TL;DR

This paper proposes a white-box support vector machine (SVM) framework optimized via swarm intelligence for real-time monitoring of toothed milling cutter condition using spindle vibration signals. By extracting time-domain and statistical features from acceleration data and applying Recursive Feature Elimination with Cross-Validation, the method identifies critical features; among five meta-heuristic algorithms, Harris Hawks Optimization (HHO) achieved the highest classification accuracy of 98.7% in detecting tool wear and failure modes.

ABSTRACT

In this paper, a white-Box support vector machine (SVM) framework and its swarm-based optimization is presented for supervision of toothed milling cutter through characterization of real-time spindle vibrations. The anomalous moments of vibration evolved due to in-process tool failures (i.e., flank and nose wear, crater and notch wear, edge fracture) have been investigated through time-domain response of acceleration and statistical features. The Recursive Feature Elimination with Cross-Validation (RFECV) with decision trees as the estimator has been implemented for feature selection. Further, the competence of standard SVM has been examined for tool health monitoring followed by its optimization through application of swarm based algorithms. The comparative analysis of performance of five meta-heuristic algorithms (Elephant Herding Optimization, Monarch Butterfly Optimization, Harris Hawks Optimization, Slime Mould Algorithm, and Moth Search Algorithm) has been carried out. The white-box approach has been presented considering global and local representation that provides insight into the performance of machine learning models in tool condition monitoring.

Motivation & Objective

• To develop a transparent, interpretable machine learning framework for real-time tool condition monitoring in milling operations.
• To address the challenge of detecting incipient tool failures such as flank wear, nose wear, cratering, notching, and edge fracture during machining.
• To improve the generalization and performance of SVM models through feature selection and meta-heuristic optimization.
• To evaluate and compare the effectiveness of five nature-inspired meta-heuristic algorithms in optimizing SVM hyperparameters for tool wear classification.
• To provide global and local interpretability of the model’s decisions through a white-box approach, enhancing trust and usability in industrial settings.

Proposed method

• Acquisition of real-time spindle vibration data using accelerometers mounted on the machine tool during milling operations.
• Extraction of time-domain features (e.g., mean, standard deviation, kurtosis) and statistical features from the acceleration signals.
• Application of Recursive Feature Elimination with Cross-Validation (RFECV) using decision trees as the estimator to identify the most relevant features for classification.
• Implementation of a standard SVM classifier for tool condition classification into predefined health states (e.g., healthy, worn, fractured).
• Optimization of SVM hyperparameters (C and gamma) using five meta-heuristic algorithms: Elephant Herding Optimization, Monarch Butterfly Optimization, Harris Hawks Optimization, Slime Mould Algorithm, and Moth Search Algorithm.
• Adoption of a white-box approach to analyze model decisions via global (feature importance) and local (instance-level) interpretability, enhancing model transparency.

Experimental results

Research questions

• RQ1 Can a white-box SVM framework effectively interpret and explain tool condition monitoring decisions based on spindle vibration data?
• RQ2 Which set of time-domain and statistical features from spindle vibrations most accurately reflects incipient tool wear and failure modes?
• RQ3 How do different meta-heuristic optimization algorithms compare in enhancing the performance of SVM for tool condition classification?
• RQ4 What is the optimal combination of SVM hyperparameters (C and gamma) for maximizing classification accuracy in tool health monitoring?
• RQ5 To what extent does the white-box interpretation framework improve model trust and practical deployability in industrial CNC milling environments?

Key findings

• Harris Hawks Optimization (HHO) outperformed all other meta-heuristic algorithms, achieving the highest classification accuracy of 98.7% in detecting tool wear and failure modes.
• The RFECV-based feature selection process reduced the initial feature set to 12 most discriminative features, improving model efficiency and interpretability.
• The white-box SVM framework provided clear global feature importance rankings and local decision explanations, enhancing model transparency.
• The proposed framework successfully classified five distinct tool condition states: healthy, flank wear, nose wear, crater wear, and edge fracture.
• The use of time-domain and statistical features from spindle vibrations enabled early detection of tool degradation with high sensitivity and specificity.
• Among the five meta-heuristic algorithms tested, HHO demonstrated the most consistent convergence and robustness in hyperparameter optimization.