LITERATURE SURVEY

Literature Survey

Tool wear monitoring is essential in modern manufacturing to maintain machining efficiency, ensure product quality, and minimize downtime. With advancements in digital manufacturing, various simulation-based, sensor-driven, and data-oriented methods have emerged for real-time tool condition monitoring.

Previous research highlights the effectiveness of vibration-based monitoring in detecting tool wear. Studies show that characteristics like RMS, kurtosis, and frequency shifts in vibration data are strong indicators of tool condition. Recent approaches integrate machine learning and FEA to enhance the accuracy of predictions. However, few studies have developed an end-to-end architecture model using MATLAB Simulink that includes sensor simulation, microcontroller interfacing, FEA, and predictive analysis within one system.

Tool condition monitoring (TCM) plays a pivotal role in ensuring process stability and product quality in advanced machining operations. Over the years, various methods combining signal processing, modal analysis, and model-based estimation have been explored to predict and monitor tool wear. The literature reveals a trend toward integrating real-time data from vibration, acoustic emission (AE), and cutting force signals with advanced analytics such as empirical mode decomposition (EMD) and machine learning for predictive maintenance.

Chi et al. (2018) proposed a real-time tool wear estimation method using vibration signal analysis through Ensemble Empirical Mode Decomposition (EEMD). By extracting modal features from vibration signals during milling, they identified specific frequency bands (notably 11.2–11.4 kHz) that showed strong correlation with tool wear. These features were then modeled using Gaussian fitting to predict tool wear trends. This approach is highly relevant to our project as it demonstrates how vibration-based modal analysis can be effectively used to extract tool wear indicators, which we aim to enhance further with FEA-based mode shape correlation and machine learning [1].

A study on fault detection system using cutting force monitoring and an observer-based estimation framework. Their method involved creating a linear dynamic model of the milling process and estimating tool wear by comparing predicted and actual force signals. Wear was inferred from trends in observer error and force signal deviations. While our project focuses on vibration rather than force signals, the concept of using dynamic system modeling and observer-based estimation aligns with our goal of integrating physical simulation (via FEA) with sensor data for predictive tool wear monitoring [2].

Recent studies have focused on improving tool condition monitoring (TCM) through advanced signal processing, simulation, and machine learning. Alonso and Salgado proposed a method using Singular Spectrum Analysis (SSA) and cluster analysis to decompose vibration signals for detecting tool wear during turning, showing that their neural network model achieved accurate and cost-effective monitoring suitable for industrial use [3]. Razika and Idriss complemented this by employing Finite Element Method (FEM) in ANSYS to analyze natural frequencies and deformation in tools with varying flank wear. Their results demonstrated that wear-induced geometry changes significantly affect vibrational behavior and tool stability [4]. Expanding the scope to machine components, Jia et al. developed a long-term monitoring system using Operational Modal Analysis (OMA) to track dynamic degradation in feed drive systems. They linked changes in modal parameters and spindle current features to mechanical wear, enhancing prediction of machine health over time [5]. Together, these works provide a foundation for hybrid, intelligent monitoring systems that address both tool and machine-level wear in modern manufacturing.

Aralikatti et al. compared tool fault diagnosis using vibration and cutting force signals with machine learning. By applying wavelet-based feature extraction and Naïve Bayes classification, they found that cutting force signals provided significantly higher accuracy (96.7%) than vibration signals (70%). This highlights the superiority of force signals for reliable tool fault detection in machining.[6] To enhance tool and bearing wear diagnostics in industrial systems, Yuan et al. [7] proposed a tool condition monitoring method using spindle motor current signals, applying Variational Mode Decomposition (VMD) for feature extraction and ensemble learning (EL) for classification. The method outperformed traditional wavelet-based techniques in both accuracy and robustness, offering a cost-effective and non-intrusive solution suitable for real processing conditions. Meanwhile, Chen et al. [8] introduced a novel framework for identifying wear parameters in hydrodynamic bearings based on Operational Modal Analysis (OMA) and on-rotor sensing (ORS). By combining stochastic subspace identification (SSI) and particle swarm optimization (PSO), their model accurately estimated wear-related shifts in modal frequencies. Together, these studies demonstrate the value of combining advanced signal processing with intelligent algorithms for real-time and accurate wear monitoring in rotating and cutting systems.