Appearance
A BTA drilling operation on a 6-metre turbine shaft begins to chatter at 3 metres depth. The operator cannot hear it — the cutting zone is 3 metres inside the workpiece, submerged in coolant at 100 bar. Within 30 seconds, the vibration amplitude grows beyond the guide pad clearance, and the pads begin to hammer against the bore wall. By 3.5 metres depth, the bore is oval by 0.15 mm and the surface finish is Ra 6.3 µm. The shaft — worth €12,000 — is scrapped. An acoustic emission sensor on the workpiece support would have detected the 703 Hz chatter frequency within 20 mm of its onset, giving the operator 180 mm of drilling to react.
Chatter and Spiralling in BTA Deep Hole Drilling
BTA deep hole drilling is inherently susceptible to dynamic instability because the cutting forces act at the end of a long, slender drill tube. The tube has low stiffness in bending and torsion, and the cutting process itself generates forces that can excite vibration.
Chatter: Self-Excited Regenerative Vibration
Chatter in BTA drilling is a self-excited vibration. The mechanism works as follows:
| Stage | Description |
|---|---|
| 1 | The cutting edge encounters a perturbation (material hardness variation, chip breakage, external vibration) |
| 2 | The perturbation causes a momentary variation in cutting force |
| 3 | The force variation excites the tool-workpiece system at its natural frequency |
| 4 | The resulting vibration leaves a wavy surface on the bore wall |
| 5 | The next cutting edge encounters this wavy surface, creating a new force variation |
| 6 | The force variation reinforces the vibration — amplitude grows with each revolution |
This regenerative loop converts a stable cutting process into an unstable one within seconds.
Spiralling: A Related Instability
Spiralling is a more severe form of dynamic disturbance that produces a characteristic multilobe bore geometry:
| Property | Stable Cutting | Chatter | Spiralling |
|---|---|---|---|
| Bore surface | Smooth, uniform | Wavy, chatter marks | Multilobe, spiral pattern |
| Roundness | Within tolerance | 0.02–0.05 mm deviation | 0.05–0.20 mm deviation |
| Surface finish Ra | 0.8–1.6 µm | 1.6–6.3 µm | 6.3–12.5 µm |
| Tool wear rate | Normal | Accelerated 2–3× | Catastrophic |
| Process stability | Stable | Marginally stable | Unstable |
Characteristic Frequencies
Research at the University of Dortmund identified two dominant frequency components in BTA chatter:
| Frequency | Association | Physical Source |
|---|---|---|
| 703 Hz | Chatter onset | Bending vibration of the drill tube at its first natural frequency |
| 1,183 Hz | Advanced chatter | Torsional vibration of the drill tube, higher mode excitation |
These frequencies appear in the vibration spectrum during unstable cutting and are absent during stable drilling. Their onset is the earliest detectable indicator of chatter.
Acoustic Emission Principles for Cutting Process Monitoring
Acoustic emission (AE) refers to the stress waves generated by the rapid release of energy within a material. In metal cutting, the primary sources of AE are:
| AE Source | Frequency Range | Physical Mechanism |
|---|---|---|
| Primary shear zone deformation | 100–300 kHz | Plastic deformation of the workpiece material ahead of the cutting edge |
| Secondary shear zone (rake face) | 200–500 kHz | Chip sliding against the tool rake face |
| Tool-chip friction | 100–400 kHz | Rubbing contact between chip and tool |
| Tool flank wear | 200–500 kHz | Rubbing of worn flank against the machined surface |
| Edge chipping and fracture | 300–600 kHz | Brittle fracture of carbide cutting edges |
| Guide pad rubbing | 100–300 kHz | Contact between guide pads and bore wall |
Why AE for Chatter Detection?
AE offers several advantages over other sensing methods for chatter detection in deep hole drilling:
| Advantage | Explanation |
|---|---|
| High frequency response | AE sensors respond to signals up to 1 MHz, capturing the onset of chatter before it becomes mechanically observable |
| Sensitivity to incipient instability | AE detects the micro-scale material deformation changes that precede macroscopic vibration |
| Immunity to low-frequency noise | The high-frequency AE band (>100 kHz) is not affected by machine tool vibration, coolant pump noise, or other low-frequency shop-floor sources |
| Non-invasive mounting | AE sensors can be mounted on the workpiece surface or tool holder without affecting the cutting process |
Sensor Types and Placement
AE Sensor Specifications for Deep Hole Drilling
| Parameter | Recommendation |
|---|---|
| Sensor type | Wideband piezoelectric AE sensor |
| Frequency range | 100 kHz – 1 MHz |
| Resonant frequency | 150 kHz (broadband) or 300 kHz (resonant for high sensitivity) |
| Pre-amplifier | 40–60 dB gain, built-in bandpass filter |
| Sampling rate | ≥ 2 MHz for full AE bandwidth, ≥ 20 kHz for vibration monitoring |
Sensor Placement Options
| Location | Signal Attenuation | Sensitivity to Chatter | Practical Issues |
|---|---|---|---|
| Workpiece surface (near bore entry) | Low (short path to cutting zone) | High | Must survive coolant splash; cable routing |
| Workpiece tailstock centre | Moderate | Moderate | Accessible during drilling; signal passes through workpiece material |
| Drill tube (near drive end) | High (long path through tube) | Low-Moderate | Sensor rotating with tube; wireless telemetry required |
| Machine spindle housing | Very high | Low | Convenient but distant from cutting zone |
| Coolant return line | N/A (fluid-borne) | Low (indirect) | Detects chip breakage events more than chatter |
Tip: The workpiece surface near the bore entry is the optimal sensor location for production deep hole drilling. The signal path to the cutting zone is short and direct, and the sensor does not rotate. A magnetic-mount AE sensor can be repositioned between workpieces without permanent installation.
Alternative and Complementary Sensors
| Sensor Type | Signal Monitored | Sampling Rate | Chatter Detection Capability |
|---|---|---|---|
| Accelerometer | Drill tube vibration | 10–20 kHz | Good — detects vibration directly |
| Torque transducer | Cutting torque | 5–20 kHz | Very good — the standard method in published research |
| Spindle current sensor | Motor power | 1–10 kHz | Moderate — low-pass filtered by drive system |
| Coolant pressure transducer | Coolant pressure | 1–5 kHz | Low — indirect, affected by other process variables |
| Microphone | Airborne sound | 20–100 kHz | Low — masked by coolant and shop noise |
Research by Weihs, Theis, and Messaoud at the University of Dortmund established torque measurement at 20 kHz sampling as the reference method for BTA chatter detection. The torque signal is processed through nonlinear time series models (exponential autoregressive models) and multivariate control charts to detect the transition from stable to unstable cutting.
Signal Processing Methods
Time-Domain Features
| Feature | Calculation | Chatter Indication |
|---|---|---|
| RMS amplitude | √(mean of squared signal) | Increases with chatter onset |
| Signal variance | Mean of squared deviations | Increases with instability |
| Zero-crossing rate | Crossings per unit time | Changes with frequency content shift |
| Crest factor | Peak / RMS | Spikes during impact events |
| Autocorrelation decay | Correlation at increasing lag | Faster decay indicates random vibration |
Frequency-Domain Analysis
Fast Fourier Transform (FFT) of the AE signal reveals the characteristic frequencies:
| Frequency Band | Source During Stable Cutting | Source During Chatter |
|---|---|---|
| < 100 Hz | Spindle rotation (2–20 Hz), feed variation | No significant change |
| 100–500 Hz | Machine structural vibration | Chatter harmonics |
| 500–800 Hz | — | 703 Hz fundamental chatter frequency |
| 800–1,500 Hz | — | 1,183 Hz chatter harmonic, torsional mode |
| > 1,500 Hz | Cutting edge engagement harmonics | Broadband energy increase |
Time-Frequency Analysis
Because chatter is a transient phenomenon — the transition from stable to unstable cutting occurs over seconds — time-frequency methods are essential:
| Method | Advantage for Chatter Detection |
|---|---|
| Short-time Fourier transform (STFT) | Simple implementation, clear visualisation of onset timing |
| Wavelet transform | Good time resolution at high frequencies, good frequency resolution at low frequencies |
| Hilbert-Huang transform (HHT) | Adaptive basis, excellent for non-stationary signals |
| Wigner-Ville distribution | Highest resolution but cross-term interference |
Nonlinear Time Series Modelling
The University of Dortmund group used exponential autoregressive (ExpAR) models to characterise the BTA drilling process:
| Model Component | Purpose |
|---|---|
| Linear AR terms | Capture the stable cutting dynamics |
| Exponential nonlinear term | Model the transition to instability |
| Residual analysis | Detect deviations from stable model behaviour |
| Control chart on residuals | Statistical detection of process change |
The ExpAR model is fitted to data from stable cutting. When chatter begins, the model residuals increase, and the control chart signals an out-of-control condition. This approach detects chatter within 0.1–0.3 seconds of onset.
Multivariate Control Charts for Chatter Detection
Multivariate control charts (Hotelling T², MEWMA, rMEWMA) are applied to features extracted from the AE or torque signal:
| Chart Type | Sensitivity | Robustness | Typical Detection Delay |
|---|---|---|---|
| Hotelling T² | Moderate | Moderate | 0.5–1.0 seconds |
| MEWMA | High | Low (sensitive to outliers) | 0.1–0.3 seconds |
| rMEWMA (robust) | High | High | 0.2–0.5 seconds |
The rMEWMA chart is recommended for production implementation because it combines high sensitivity to chatter onset with robustness against the occasional outliers caused by chip breakage events.
Frequency Signatures in Practice
The characteristic 703 Hz and 1,183 Hz frequencies are consistent across different BTA drilling setups because they are determined by the drill tube geometry:
| Drill Tube Parameter | Effect on Chatter Frequency |
|---|---|
| Tube length | Longer tubes → lower natural frequencies |
| Tube outer diameter | Larger diameter → higher stiffness → higher frequencies |
| Tube wall thickness | Thicker wall → higher stiffness → higher frequencies |
| Coolant pressure | Pressure stiffens the tube slightly → small frequency increase |
| Workpiece material | Higher cutting forces → lower stability threshold |
Note: Although the published research identifies 703 Hz and 1,183 Hz as characteristic frequencies for specific BTA setups, these values should be determined experimentally for each unique drilling configuration. A pre-production stability test — running the BTA head through a test workpiece at increasing feed rates while monitoring the AE spectrum — identifies the chatter frequencies for that specific tool-workpiece combination.
Comparison of AE with Alternative Sensing Methods
| Method | Sensitivity to Chatter Onset | Detection Delay | Cost | Suitability for Production |
|---|---|---|---|---|
| Acoustic emission (workpiece-mounted) | Very high | < 0.1 s | €3,000–€8,000 | Excellent |
| Torque measurement (20 kHz) | High | 0.1–0.3 s | €5,000–€15,000 | Good (requires modified machine) |
| Accelerometer (drill tube) | High | 0.2–0.5 s | €500–€2,000 | Moderate (rotating sensor) |
| Spindle current (1 kHz) | Low-Moderate | 1–5 s | €200–€500 | Good (non-invasive) |
| Coolant pressure (1 kHz) | Low | 2–10 s | €500–€2,000 | Good (non-invasive) |
| Microphone (airborne) | Low | Variable | €200–€1,000 | Poor (noise contamination) |
AE monitoring provides the earliest possible detection of chatter onset, before the vibration amplitude has grown large enough to cause bore surface damage. The combination of AE with torque monitoring provides the most robust detection system.
Industrial Implementation
System Architecture
A production-ready AE-based chatter detection system consists of:
| Component | Function |
|---|---|
| AE sensor (workpiece-mounted) | Captures stress waves from the cutting zone |
| Preamplifier (40–60 dB) | Amplifies the AE signal for transmission |
| Data acquisition (≥ 2 MHz) | Digitises the AE signal |
| Signal processor | FFT, feature extraction, model prediction |
| Control chart engine | rMEWMA or similar for statistical detection |
| Machine interface | Digital output for stop/warning signal |
| HMI display | Real-time chatter indicator for operator |
Implementation Workflow
- Pre-production stability test — Drill a test workpiece at incremental feed rates while recording AE and torque. Identify the chatter threshold and characteristic frequencies.
- Model fitting — Fit the ExpAR model or train the classifier using data from stable cutting conditions.
- Control limit establishment — Set rMEWMA control limits using Phase I data from stable drilling.
- Production monitoring — Run the system continuously during production drilling.
- Response procedure — Define operator actions for warning (reduce feed, check tool) and alarm (stop drill, retract, inspect) signals.
Response to Chatter Detection
| Signal | Indication | Operator Action |
|---|---|---|
| Warning level 1 | Early chatter onset, 703 Hz component emerging | Reduce feed rate by 20%; monitor signal |
| Warning level 2 | Chatter established, amplitude growing | Stop feed; continue spindle rotation for 5 s; resume at reduced feed |
| Alarm level 1 | Strong chatter, risk of bore damage | Stop drilling immediately; retract tool; inspect bore |
| Alarm level 2 | Spiralling detected | Stop drilling immediately; workpiece likely scrap |
Troubleshooting
| Problem | Likely Cause | Corrective Action |
|---|---|---|
| No AE signal from sensor | Sensor not coupled to workpiece surface | Apply coupling gel; verify magnetic hold; clean mounting surface |
| False chatter alarms | Chip breakage events producing AE bursts | Increase control limit; apply signal filtering for chip events |
| Chatter not detected before bore damage | Sensor too far from cutting zone; signal attenuated | Move sensor closer to bore entry; increase preamplifier gain |
| Inconsistent frequency signature | Variable cutting conditions between workpieces | Normalise frequency features by cutting parameters |
| High background AE noise | Coolant cavitation or pump noise | Apply high-pass filter (>150 kHz); use differential sensor |
| rMEWMA false positives | Model fitted to limited training data | Increase Phase I data collection to 30+ stable cycles |
FAQ
What is chatter in BTA deep hole drilling?
Chatter is a self-excited regenerative vibration that occurs when the cutting process excites the natural frequency of the tool-workpiece system. In BTA drilling, the long slender drill tube has low stiffness, making it susceptible to chatter. The vibration grows rapidly and damages the bore surface, typically producing a wavy surface finish and oversize diameter.
What is spiralling in deep hole drilling?
Spiralling is a severe form of dynamic instability in BTA drilling where the drill tube vibrates in a mode that produces a multilobe bore cross-section. The lobes form a spiral pattern along the bore length. Spiralling causes ovality that can exceed the tolerance by 10× and typically requires scrapping the workpiece.
What frequencies indicate chatter in BTA drilling?
Published research identifies 703 Hz and 1,183 Hz as characteristic chatter frequencies for BTA drilling. The 703 Hz component corresponds to the first bending natural frequency of the drill tube; 1,183 Hz corresponds to a torsional or higher bending mode. These frequencies should be verified experimentally for each drilling configuration.
How does acoustic emission detect chatter?
AE sensors mounted on the workpiece surface capture the stress waves generated by the cutting process. The high-frequency AE signal (>100 kHz) changes character when chatter begins — specific frequency components (e.g., 703 Hz) appear or increase in amplitude. Signal processing methods (FFT, time-frequency analysis, nonlinear time series models) identify these changes, and multivariate control charts detect the transition from stable to unstable cutting.
Where should AE sensors be placed for deep hole drilling?
The optimal location is on the workpiece surface near the bore entry. This provides a short, direct signal path from the cutting zone to the sensor. Magnetic-mount AE sensors can be repositioned between workpieces. The sensor should be coupled to the workpiece surface with a thin layer of coupling grease.
How early can AE detect chatter compared to other methods?
AE detects chatter within 0.1 seconds of onset, before the vibration amplitude has grown large enough to affect the bore surface. Torque measurement detects chatter at 0.1–0.3 seconds. Spindle current monitoring detects it at 1–5 seconds, by which time bore damage may already have occurred.
What signal processing is needed for AE-based chatter detection?
A combination of frequency-domain analysis (FFT to identify characteristic frequencies), time-frequency analysis (STFT or wavelet transform to track onset timing), and statistical process control (multivariate control charts on extracted features) provides the most reliable detection.
What are the characteristic frequencies of BTA chatter?
The primary chatter frequency is approximately 703 Hz, corresponding to the bending natural frequency of the drill tube. A secondary frequency at approximately 1,183 Hz appears during advanced chatter. These values are specific to the drill tube geometry and should be determined experimentally for each setup.
Can AE distinguish between chatter and normal chip breakage?
Yes. Normal chip breakage produces short, broadband AE bursts at irregular intervals. Chatter produces a sustained increase in narrowband energy at specific frequencies (703 Hz, 1,183 Hz). Frequency-domain analysis and time-frequency analysis distinguish between the two. Multivariate control charts on frequency-band energy features provide robust discrimination.
What is the economic case for AE chatter detection?
A single scrapped workpiece in deep hole drilling costs €5,000–€15,000. The AE monitoring system costs €3,000–€8,000 per machine. In a facility where chatter events occur 1–5 times per year, the payback period is under 12 months. In operations with marginal stability (e.g., drilling at the limits of the process capability), the payback can be under 3 months.
Conclusion
Acoustic emission monitoring provides the earliest possible detection of chatter and spiralling in BTA deep hole drilling. The high-frequency AE signal (100–500 kHz) captures the onset of regenerative vibration within 0.1 seconds — before the vibration amplitude grows large enough to damage the bore surface. The characteristic frequencies of 703 Hz and 1,183 Hz, identified through extensive research at the University of Dortmund, provide specific targets for frequency-domain monitoring. When combined with multivariate control charts (particularly the robust rMEWMA) and nonlinear time series models, AE-based chatter detection delivers reliable process monitoring with low false alarm rates. The three engineering priorities for AE chatter detection in deep hole drilling are: placing the AE sensor as close as possible to the bore entry on the workpiece surface for maximum signal-to-noise ratio, characterising the chatter frequencies for each specific tool-workpiece configuration during pre-production testing, and implementing multivariate control chart monitoring with response procedures that allow the operator to take corrective action (reduce feed, retract and restart) before the bore surface is damaged.