UNIVERSITY OF VICTORIA · MASc RESEARCH

Operational Modal Analysis
for Chatter Prediction in Turning

Estimating machining stability from in-process vibration measurements by identifying dominant structural poles, modal damping, and the approach to regenerative instability.

View Publication
Published Research

From graduate research to peer-reviewed publication

The work was published in Mechanical Systems and Signal Processing as “Estimation of vibration stability in turning using operational modal analysis.”

MSSP Mechanical Systems and Signal Processing
2019 Publication year
OMA Operational modal analysis
Chatter Stability prediction
Research Problem

Detecting instability before chatter fully develops

Conventional chatter-monitoring methods commonly detect instability after vibration amplitudes have already grown and chatter marks may have formed on the workpiece.

The objective of this research was different: estimate the stability margin while the cutting process is still stable by identifying the dynamics of the turning system directly from vibration measurements obtained during machining.

Research objective

  • Measure machining vibration during stable cutting
  • Identify dominant structural poles
  • Estimate modal frequency and damping
  • Track damping as stability decreases
  • Detect the approach to regenerative chatter
Experimental System

Instrumented orthogonal-turning experiments

The experimental system combined controlled machining conditions with multi-channel vibration acquisition and independent acoustic monitoring.

Experimental turning setup inside CNC machine
Experimental turning setup used to evaluate process stability under different spindle speeds and widths of cut.

Machining configuration

  • Aluminum 6061 workpiece
  • 38 mm tool overhang
  • 38 mm diameter workpiece
  • Orthogonal turning configuration
  • Multiple spindle-speed / width-of-cut conditions

Measurement system

  • Four PCB accelerometers
  • Feed and cutting directions
  • NI data-acquisition hardware
  • 16,384 Hz vibration sampling
  • Microphone for independent chatter verification
Instrumentation

Multi-point vibration measurement

Accelerometers were positioned at two locations in each principal tool-vibration direction, providing four measured degrees of freedom for modal identification.

This multi-channel configuration enabled direct and cross-correlation analysis between vibration signals measured at different locations and directions.

Accelerometer positions on the turning tool holder
Sensor layout showing the four measured degrees of freedom used for vibration acquisition and system identification.
Close-up of turning tool and cylindrical workpiece
Close-up of the orthogonal-turning configuration used for the machining experiments.
Regenerative Chatter

Instability leaves a permanent machining signature

Regenerative chatter arises from delayed feedback between the current cutting pass and surface waviness generated during the previous revolution.

Once unstable vibration fully develops, the resulting oscillation can degrade surface quality and leave visible chatter marks on the workpiece.

Machined workpiece showing chatter marks
Surface pattern produced under unstable cutting conditions.
Dynamics & Control Foundations

A machining problem formulated as a dynamic-system stability problem

Although the application was machining dynamics, the underlying framework was fundamentally a dynamic-systems and stability problem.

Closed-loop numerical simulation model of self-excited turning vibration
Numerical simulation framework representing modal dynamics, regenerative delayed feedback, excitation, and measured vibration response.
MODELING

State-space dynamics

Multi-degree-of-freedom tool dynamics were represented in modal and state-space form for numerical simulation.

FEEDBACK

Time-delay dynamics

Regenerative cutting forces were represented as delayed feedback linking the current tool motion to vibration from the previous workpiece revolution.

STABILITY

Pole and eigenvalue analysis

Stability was evaluated through the dominant system poles and their corresponding damping ratios.

Connection to later robotics work

Although the application was machining dynamics, the analytical framework relied on many of the same foundations that later became central to my robotics work: state-space modeling, feedback dynamics, numerical simulation, eigenvalue-based stability analysis, system identification, and sensor-driven estimation.

Signal Processing

Extracting structural dynamics from machining vibration

Strong spindle harmonics coexist with structural vibration during turning, making direct modal identification significantly more difficult than in controlled impact testing.

Measured machining signal and power spectral density
Measured process signal and frequency-domain representation. Repeated spindle harmonics appear throughout the spectrum while structural vibration develops in their vicinity.
01 Record Vibration
→
02 Detrend
→
03 PSD Analysis
→
04 Band-Pass Filter
→
05 Correlation Function
→
06 Pole Identification
System Identification

Separating structural modes from spindle harmonics

Harmonic excitation from spindle rotation can dominate the measured response and interfere with conventional output-only modal identification.

Modified LSCE stabilization diagram for structural pole identification
Modified LSCE identification treats known spindle harmonics as undamped poles, enabling the nearby damped structural mode to be identified separately.

TDPR

Time Domain Poly Reference identification was used successfully in numerical simulations but became less reliable when strong experimental harmonics were present.

Modified LSCE

Known harmonic frequencies were incorporated explicitly as undamped poles, allowing nearby structural poles and modal damping to be estimated more robustly.

Experimental Result

Damping decreased as the process approached instability

The central result was not simply detecting chatter. The identified structural damping provided a quantitative indication of how close the process was to its stability boundary.

Spindle Speed Width of Cut Identified Mode Damping Condition
7500 rpm 2.03 mm ~2476 Hz ~0.13% Stable
7500 rpm 2.54 mm ~2477 Hz ~0.066% Near stability boundary
7500 rpm 3.05 mm Chatter — Unstable
Experimental stability map showing damping ratio and stable and unstable turning conditions
Experimental stability map. Stable operating points are associated with identified modal damping, while unstable conditions are marked by chatter. The damping decreases as the operating condition approaches the stability boundary.
Engineering Interpretation

Stability can be treated as an estimated dynamic state

Traditional chatter monitoring focuses on identifying vibration after instability has already developed. This work instead estimated the dynamics of the system while machining remained stable.

As the stability boundary was approached, the damping of the dominant structural pole decreased. This provided a physically meaningful numerical indicator of declining stability before the vibration became fully unstable.

Key findings

  • Structural poles can be identified from process vibration
  • Modal damping decreases near the stability boundary
  • Modified LSCE improves identification near spindle harmonics
  • Stable-process vibration contains information about stability margin
  • OMA can provide an early indicator of chatter onset
Limitations

Identification remains difficult under some operating conditions

The method was most effective when the process was close enough to the stability boundary for the structural mode to become observable in the measured vibration spectrum.

Highly stable cuts and operating conditions where a structural frequency coincided with a spindle harmonic could suppress the structural mode and prevent reliable identification.

Future work

  • Automated modal identification
  • Reduced computational time
  • Real-time implementation
  • Improved robustness near harmonic coincidence
My Research

Modeling, experimentation, signal analysis, and validation

This work formed my MASc research at the University of Victoria and combined analytical modeling, numerical simulation, experimental modal analysis, machining experiments, sensor instrumentation, MATLAB-based signal processing, and system identification.

The research strengthened the dynamic-systems and numerical analysis foundation that later carried directly into my work in robotics, autonomous systems, estimation, and control.

Methods & tools

  • MATLAB
  • Dynamic-system modeling
  • State-space simulation
  • Operational modal analysis
  • Experimental modal analysis
  • System identification
  • Spectral / correlation analysis
  • Accelerometers & DAQ
  • CNC machining experiments
PUBLICATION

Estimation of vibration stability in turning using operational modal analysis

S. Kim · K. Ahmadi
Mechanical Systems and Signal Processing, 2019

View Publication
← Back to University of Victoria Research