Exploring Torsional Vibration Through MATLAB Simulation: An Academic Approach

Torsional vibration in drill strings is a recognized problem in oil and gas drilling that is difficult to measure directly downhole and challenging to predict with simple rules. A Master’s thesis explored how mathematical simulation might help understand these oscillations—not by solving the problem definitively, but by building a theoretical model to investigate the mechanisms.

This post examines that exploratory work: how a two-degree-of-freedom mathematical model, implemented in MATLAB, can be used to investigate stick-slip behavior and explore potential mitigation approaches. The emphasis is on methodology and investigation, not on solutions proven to work in practice.

Part 1: Why Simulation Matters in Drilling

The Challenge

When a drill string rotates at depth, torsional oscillations can occur—the string twisting back and forth around its axis. These oscillations are difficult to measure directly downhole and are not well understood from first principles. In severe cases, stick-slip behavior (alternating sticking and slipping of the bit) can develop, leading to:

• Cyclic fatigue stress in the drill string
• Tool wear acceleration
• Potential stuck pipe situations
• Reduced drilling efficiency

Measurement Challenge: Direct downhole measurement of torsional vibration at 3,000+ meters is technically difficult. Sensors must withstand extreme pressure, temperature, and corrosive conditions. Surface measurements (torque at the top drive) are indirect proxies.

The Research Question

Rather than relying on experience-based “rules of thumb” that vary by region and operator, could a mathematical model help understand the mechanisms causing stick-slip? The MSOG thesis investigated this question by:

1. Building a simplified mathematical model of the drill string
2. Implementing it in MATLAB for numerical exploration
3. Conducting sensitivity analyses to identify critical parameters
4. Testing a proposed mechanical mitigation device (AST) in simulation

Important: This is exploratory research aimed at understanding mechanisms, not a finished solution. The approach is theoretical, and many questions remain unanswered.

Part 2: The Mathematical Model (Two Degree-of-Freedom System)

Model Simplification Strategy

The actual drill string is a complex structure with many components. The thesis simplified it to a two degree-of-freedom (2-DOF) torsional pendulum for mathematical tractability:

Surface Motor (Torque: Tm)
    ↓
[Rotary Table] ← Inertia: Jr = 2122 kg⋅m²
    ↓ (Spring k, Damper c)
[Drill Pipe Assembly] (lumped as massless spring)
    ↓
[BHA + Bit] ← Inertia: Jb = 471.97 kg⋅m²
    ↓
Rock Formation (Reaction torque: Tb)

Model Design Choices (from Chapter 3):

• Two inertias: Rotary table at top, BHA at bottom (everything between is a spring)
• Rotary table inertia (Jr = 2122 kg⋅m²): Top drive system’s rotational mass
• BHA inertia (Jb = 471.97 kg⋅m²): Bottom hole assembly (collars, bit, tools treated as one rigid body)
• Spring constant (k = 861.53 N⋅m/rad): Represents torsional compliance of all drill pipe between
• Damping (c = 425 N⋅m⋅s/rad): Lumped viscous damping coefficient

Acknowledgment of Simplifications: This 2-DOF model makes many assumptions (linear elasticity, constant properties, ignores lateral/axial coupling, treats distributed mass as lumped) that the real drill string doesn’t satisfy. Whether these simplifications are acceptable for understanding stick-slip is an open question that would require experimental validation to answer.

The Equations of Motion

The fundamental equations governing the system are:

For the rotary table:

Jᵣ·φ̈ᵣ = Tₘ - Tₐ - c(φ̇ᵣ - φ̇ᵦ) - k(φᵣ - φᵦ)     (3.4)

For the BHA and bit:

Jᵦ·φ̈ᵦ = -Tᵦ - c(φ̇ᵦ - φ̇ᵣ) - k(φᵦ - φᵣ)     (3.5)

Where:

• φ̈, φ̇, φ = angular acceleration, velocity, displacement
• Tₘ = input motor torque
• Tₐ = damping torque at rotary table (neglected for simplicity)
• Tᵦ = torque from bit-rock interaction (nonlinear, depends on friction)
• c(φ̇ᵣ - φ̇ᵦ) = viscous damping torque proportional to relative velocity
• k(φᵣ - φᵦ) = spring torque proportional to relative displacement

The Friction Torque Model (Proposed Representation)

A critical part of the model is how friction at the bit is represented. The thesis proposes a friction model with three regimes. This is a simplification of real friction behavior, which is complex and not fully understood. The proposed model:

1. Stick Phase (bit not rotating):

Tᶠᵦ = Tₑᵦ = c(φ̇ᵣ - φ̇ᵦ) + k(φᵣ - φᵦ) - cᵦφ̇ᵦ     (3.13)

The friction adjusts to match the external torque.

2. Stick-to-Slip Transition:

Tᶠᵦ = Tₛᵦ·sgn(Tₑᵦ)     (3.9)

When external torque exceeds static friction Tₛᵦ = Rᵦ·Wᵦ·μₛᵦ ≈ 5227 Nm, bit begins slipping.

3. Slip Phase (bit rotating):

Tᶠᵦ = Rᵦ·Wᵦ·[μcᵦ + (μₛᵦ - μcᵦ)·e^(-γᵦ|φ̇ᵦ|)]·sgn(φ̇ᵦ)     (3.12)

Coulomb friction torque Tcᵦ ≈ 3300 Nm decays exponentially with bit velocity.

The exponential decay term (equation 3.12) was proposed as a way to represent the transition from static to kinetic friction. Whether this exponential model accurately represents real bit-rock friction is unknown—the thesis cites laboratory friction studies, but actual drilling friction under high pressure, temperature, and with real formations remains not well characterized.

Implementation: Converting to State-Space Form

To implement this model numerically in MATLAB, the second-order differential equations must be converted to a system of first-order equations (ODE45 integrator requirement):

x₁ = φ̇ᵣ  (rotary table velocity)
x₂ = φᵣ - φᵦ  (relative displacement - the key variable)
x₃ = φ̇ᵦ  (bit velocity)

The state equations become:

ẋ₁ = (1/Jᵣ)[Tₘ + c(x₃ - x₁) - k·x₂]     (3.16a)

ẋ₂ = x₁ - x₃                             (3.16b)

ẋ₃ = (1/Jᵦ)[-Tᶠᵦ + c(x₁ - x₃) - cᵦ·x₃ + k·x₂]     (3.16c)

This is the model MATLAB integrates numerically using ODE45 with adaptive time-stepping.

The Actual MATLAB Implementation

The MSOG thesis implements this model in MATLAB using the ODE45 solver, which integrates the system of first-order differential equations with adaptive time-stepping.

Simulation Parameters (from actual thesis):

Jr = 2122 kg⋅m²           % Rotary table inertia
Jb = 471.9698 kg⋅m²       % BHA inertia
k = 861.5336 N⋅m/rad      % Spring stiffness
c = 425 N⋅m⋅s/rad         % Damping coefficient
cb = 100-150 N⋅m⋅s/rad    % Bit damping (varied)
Tm = 4100 N⋅m             % Drive torque
Wb = 42000 N              % Weight on bit
Rb = 0.155575 m           % Bit radius
μcb = 0.5                 % Coulomb friction coefficient
μsb = 0.8                 % Static friction coefficient
γb = 0.9                  % Exponential decay rate

Key Integrator Code from Appendix B:

% Friction torque calculation with three regimes
Teb = c*(x(1)-x(3)) + k*x(2) - cb*x(3);  % External torque
Tsb = Rb*Wb*musb;                         % Static friction threshold

if (abs(x(3)) >= Dv)  % Slip phase (bit rotating)
    % Exponential friction model
    Tfb = Rb*Wb*(mucb + (musb-mucb)*exp(-gamb*abs(x(3))))*sign(x(3));

elseif (abs(Teb) > Tsb)  % Stick-to-slip transition
    Tfb = Tsb*sign(Teb);

else  % Stick phase (bit stationary)
    Tfb = Teb;
    dx(3) = 0;  % Bit velocity remains zero
end

% State derivatives (equations 3.16)
dx(1) = (1/Jr)*(Tm + c*(x(3)-x(1)) - k*x(2));
dx(2) = x(1) - x(3);
dx(3) = (1/Jb)*((-cb*x(3)) + c*(x(1)-x(3)) + k*x(2) - Tfb);

The solver runs for 100 seconds of simulated drilling time, automatically adjusting step size to maintain accuracy (RelTol: 10⁻¹⁰).

Part 3: Simulation Outcomes (Model Predictions)

Three Predicted Behaviors

The thesis’s MATLAB model produced three distinct behavioral patterns when parameters were varied. These are predictions from the mathematical model, not observations from real drilling or laboratory experiments. What the model predicts and what actually occurs downhole may differ due to the simplifications discussed above:

Regime 1: Transition to Equilibrium

Scenario: Tm = 4100 N⋅m, Wb = 41500 N, cb = 150 N⋅m⋅s/rad

Initial conditions (all parameters at rest):
x(0) = [0, 0, 0]T    (no motion)

Model prediction:

• The drill string “winds up” before the bit starts rotating
• Angular displacement reaches ~5.5 rad before bit moves
• Bit oscillates 8 times with decreasing amplitude
• After ~30 seconds, system settles to equilibrium with both rotating together at ~5.7 rad/s
• Permanent twist remains: ~4.75 rad (top always ahead of bottom)

Interpretation within the model: The torsional spring stores and releases energy. Oscillations decay due to the damping coefficient. Whether this transient behavior matches real startup dynamics is unknown and would require experimental observation to verify.

Regime 2: Permanently Stuck Bit

Scenario: Tm = 4100 N⋅m, Wb = 41900 N, cb = 150 N⋅m⋅s/rad

Critical observation: Only 400 N difference in WOB causes catastrophic change

Model prediction:

• Bit starts rotating initially but then jams completely after 38 seconds
• Friction torque (Tfb) builds to ~4400 Nm, exceeding drive torque (Tm = 4100 Nm)
• Rotary table velocity oscillates between ±3 rad/s as energy dissipates
• System eventually stalls completely

Significance within model: A small 400 N increase in WOB (from 41500 to 41900 N, or 0.96% change) shifts the system from stable equilibrium to complete lockup. This extreme sensitivity is a key result of the model. Whether this sensitivity exists in real drilling is unknown and is highly dependent on the accuracy of the friction model assumptions.

Regime 3: Stick-Slip Oscillations (The Main Problem)

Scenario: Tm = 4100 N⋅m, Wb = 42000 N, cb = 100 N⋅m⋅s/rad

Model-predicted frequency: 0.21 Hz (21 cycles over 100 simulated seconds)

The thesis notes that field observations of stick-slip typically report 0.05-0.5 Hz, so 0.21 Hz falls within the observed range. However, this doesn’t validate the model—many different mechanisms could produce oscillations in this frequency range, and agreement on frequency alone doesn’t confirm the model captures real physics.

Predicted mechanics during stick-slip cycle (from the model):

STICK PHASE (duration ~0.5 seconds):
  • Bit velocity: x₃ = 0 (completely stopped)
  • Friction torque builds: Tfb → Tsb ≈ 5227 Nm
  • (Static friction coefficient μsb = 0.8)

STICK-TO-SLIP TRANSITION:
  • When Tfb exceeds Tsb, bit begins breaking free
  • Friction torque discontinuously switches to Coulomb value
  • Tcb ≈ 3300 Nm (coulomb friction coefficient μcb = 0.5)

SLIP PHASE (velocity increasing):
  • Bit accelerates to peak velocity x₃ ≈ 4.5 rad/s
  • As velocity increases, friction exponentially decays:
    Tfb = Rb·Wb·[μcb + (μsb - μcb)·e^(-γb·|ẋ₃|)]
                = 3263 Nm at start → decreases exponentially

  • The negative slope in friction-velocity curve drives instability:
    As bit speeds up, friction decreases → less braking → more speed
    As bit cools down, friction increases → more braking → less speed

RETURN TO STICK:
  • Friction torque rises above driving torque again
  • Bit suddenly decelerates back to zero
  • Cycle repeats with 0.21 Hz frequency

Model interpretation: The model predicts 21 stick-slip cycles over 100 seconds. Each cycle involves:

• Bit velocity discontinuities (jumping 0 → 4.5 rad/s and back)
• Friction torque transitions (switching between 3300 and 5227 Nm)
• Energy dissipation in the system

What this means physically: IF the model’s friction assumptions are correct, THEN these discontinuities and energy transfers would occur. But the real bit-rock interaction may not follow the proposed exponential friction model, so the actual mechanics could differ substantially.

Part 4: Sensitivity Analysis (Model Exploration)

The thesis explored how changing parameters affects the model’s behavior. These are sensitivity results from the mathematical model—they show which parameters the model is sensitive to, not necessarily which parameters matter most in real drilling. The real-world sensitivities could be quite different.

Effect 1: Drive Torque (Tm)

Tested range: 3800 to 4300 N⋅m (constant WOB, damping)

Tm = 3800 N⋅m:  Bit oscillations continue but smaller amplitude
Tm = 4100 N⋅m:  Classic stick-slip (baseline)
Tm = 4300 N⋅m:  Larger amplitude oscillations, longer stick time

Model behavior: In the model, increasing torque doesn’t simply eliminate stick-slip. The model shows a “Goldilocks zone” where torque is neither too high nor too low. Whether this applies to real drilling depends on whether the friction model is accurate. If real friction behaves differently (e.g., not exponentially), the torque sensitivity could be entirely different.

Effect 2: Spring Constant (k = torsional stiffness)

Tested range: 500 to 1000 N⋅m/rad

Actual result:

k = 500 N⋅m/rad:   Stick-slip SUPPRESSED → stable drilling achieved
k = 861 N⋅m/rad:   Classic stick-slip (baseline configuration)
k = 1000 N⋅m/rad:  Stick-slip AMPLIFIED → higher frequency (0.24 Hz vs 0.21 Hz)

Model prediction: Lower spring stiffness suppresses stick-slip in the model. Physically, k = G·Jp/L, so:

• Lower stiffness = longer drill string or smaller diameter
• In the model: Slower spring restoring torque = bit has more time to escape stick phase before friction rebuilds

Caveat: This prediction depends critically on the exponential friction model. If real friction has a different velocity dependence, the stiffness sensitivity could reverse or disappear entirely.

Effect 3: Viscous Damping (cb)

Tested range: 50 to 150 N⋅m⋅s/rad

Actual results:

cb = 50 N⋅m⋅s/rad:    Stick-slip ELIMINATED → stable drilling (high bit velocity)
cb = 100 N⋅m⋅s/rad:   Classic stick-slip (baseline)
cb = 150 N⋅m⋅s/rad:   Bit PERMANENTLY STUCK (damping too high, bit can't escape)

Model behavior: The model shows a nonlinear damping effect—there’s an optimal window (50-100 N⋅m⋅s/rad). Too little and oscillations occur; too much and the system stalls.

Note: This nonlinear damping effect is a result of the specific friction model used. Different friction models might show damping effects that follow different rules. The physical relationship between mud viscosity and effective damping in real drilling is complex and not well characterized.

Effect 4: Weight on Bit (Wb)

Tested range: 41500 to 42500 N (300 N variation)

Actual results:

Wb = 41500 N:  Transient response only, converges to equilibrium (stable)
Wb = 42000 N:  Classic stick-slip oscillations (baseline)
Wb = 42500 N:  Bit STALLS after few oscillations (failure mode)

Model sensitivity: The model shows extreme sensitivity to WOB. A 2.4% weight change (400 N out of ~42000 N) causes transition from oscillating to completely stuck.

In the model: WOB directly scales friction torque: Tfb = Rb·Wb·μ. Higher friction makes it harder for the motor to maintain bit rotation.

Reality check: This extreme sensitivity is concerning from a practical standpoint—real drilling would need to be remarkably well-controlled to operate in such a narrow window. Either (1) real drilling is more robust than this model suggests, or (2) real drilling does operate on a knife’s edge and stick-slip is indeed barely avoidable.

Effect 5: Friction Coefficients (μcb vs μsb)

Tested variations: Static (μsb) and Coulomb (μcb) friction coefficients

Key finding—the gap matters:

Baseline (μsb = 0.8, μcb = 0.5):  Stick-slip (0.21 Hz)

If coefficients converge (gap closes):
  μsb = 0.8, μcb = 0.6 → System tends toward STICKING

If coefficients diverge (gap widens):
  μsb = 0.8, μcb = 0.4 → STABLE drilling at higher bit velocity

Model mechanism: The model’s instability depends on the friction discontinuity between static and Coulomb values. A large gap allows faster transitions and more stable operation.

Limitation: This result depends entirely on the model’s assumption that friction has discrete static/Coulomb regimes. Real friction may transition more smoothly, or may not have these distinct regimes at all under high-pressure drilling conditions.

Effect 6: Friction Decay Rate (γb)

Tested range: 0.75 to 0.90

Actual results:

γb = 0.75:  Stick time ts ≈ 2.03 seconds (system barely recovers)
γb = 0.80:  Stick time ts ≈ 1.16 seconds
γb = 0.85:  Stick time ts ≈ 0.76 seconds
γb = 0.90:  Stick time ts ≈ 0.41 seconds (baseline, unstable)

γb ≤ 0.75:  Friction switches TOO SLOWLY → bit never accumulates
            enough speed to escape stick phase → PERMANENT STALL

Model result: γb controls the exponential decay rate of friction with velocity in the model. Too slow and system stalls; too fast and oscillations occur.

Critical caveat: This parameter (γb = exponential decay rate) is not directly measurable in real drilling and was fit to literature friction models. Whether the actual bit-rock friction under drilling conditions follows an exponential decay is unknown and is one of the largest unknowns in the model.

Part 5: Anti-Stick-Slip Tool (AST) - Model Exploration

The thesis also modeled a proposed mechanical device (AST - Anti-Stick-Slip Tool): a spring-loaded tool with a helical spline interface designed to absorb and manage torsional oscillations.

AST Design (from Chapter 3):

• Helical spline with angle α, precompressed spring (F₀ = 4200 N)
• Designed to provide additional compliance without slipping initially

Model Predictions for AST

Using the same parameters that produced stick-slip without AST:

Tm = 4100 N⋅m, Wb = 42000 N, cb = 100 N⋅m⋅s/rad

Model prediction with AST (α = 10°):

• First 10 seconds: Transient response as system settles
• After 10 seconds: Model predicts stable operation with both surfaces rotating together
• No predicted stick-slip oscillations

Within the model: The spring provides additional compliance, altering the effective spring constant and changing system dynamics. The model predicts this suppresses the oscillations.

Spline angle effects (model predictions):

α = 10°:   Model shows fastest convergence to stability
α = 45°:   Model shows slower stability achievement
α = 90°:   Model shows no torque transmission → system stalls

Important Caveats

The AST results are entirely simulation-based. Key unknowns:

• Does the actual spring-spline interface behave as modeled (linear spring, no friction losses)?
• How does high-pressure drilling affect the spring properties?
• Would stick-slip mechanisms in real drilling respond to this additional compliance?
• Has this device been tested on any actual well or test rig?

The thesis does not include experimental validation of the AST concept. It remains a theoretical proposal that showed promise in the mathematical model.

Part 6: From Simulation Results to Drilling Decisions

Critical Disclaimer: This Is Academic Exploration, Not Operational Procedure

The following is NOT an operational procedure. These sections describe theoretical results from an academic two-DOF mathematical model. No testing was done to validate whether these predictions apply to real drilling. No field work, no test rig experiments, no comparison to actual wells.

The MSOG thesis successfully demonstrated that mathematical simulation can explore mechanisms and predict sensitivity patterns. But simulation validation is fundamentally different from operational validation. Between this model and a real drilling decision lie many steps:

• Does the friction model match real bit-rock interactions?
• Do the simplified equations capture actual drill string dynamics?
• Would real drilling respond to these predicted parameter changes?

Before implementing any of the following recommendations in actual drilling, field validation would be essential.

Decision Flow in the Rig

1. Plan the well section
   ↓
2. Define drill string components
   ↓
3. Run simulation (eigenvalue analysis)
   ↓
4. Calculate critical speeds
   ↓
5. Create operational envelope (safe RPM windows)
   ↓
6. Brief drilling crew on parameter limits
   ↓
7. During drilling: monitor torque and RPM for anomalies
   ↓
8. If vibration observed: adjust parameters within safe window
   ↓
9. Post-well: compare actual performance with predictions
   (validate model for next section)

Red Flags the Simulation Warns About

Warning 1: Critical speeds too close together

Mode 1: 738 RPM
Mode 2: 742 RPM  ← Too close! Hard to avoid both
       Gap = 4 RPM

Action: Change drill string configuration to separate natural frequencies

Warning 2: Low damping at critical speeds

Damping ratio ζ = 0.02 (2%)  ← very low
                  ↓
Magnification factor Q = 1/(2ζ) = 25×

Action: Select heavier BHA tools, increase formation contact, reduce speed in that region

Warning 3: Multiple natural frequencies in operating range

Safe operations: 200-600 RPM and 700-1200 RPM
Available modes: 1st at 550, 2nd at 750 RPM
                 ↓
No continuous safe window

Action: Reduce drill string length, change well design, or accept operational risk with careful speed ramps

Part 7: Theoretical Operational Guidelines (From Simulation)

Important: These Are Model Predictions, Not Validated Procedures

The information in this section represents what the MSOG thesis model predicts would happen if its assumptions are correct. These are not field-tested guidelines and should not be directly implemented in actual drilling without validation.

The thesis demonstrates that simulation can identify critical parameters and predict operational windows. But the gap between “the model predicts this” and “this will work on a real well” is large and requires experimental evidence to bridge.

The MSOG Thesis Drill String Configuration

100 drill pipes:    5" OD × 4.408" ID, 9m each
2 drill collars:    6.5" OD × 2.25" ID, 18.6m each
1 stabilizer:       12.25" OD × 6.5" ID, 1.5m
1 HWDP:             6.5" OD, 9.3m
1 bit:              12.25" OD
────────────────────────────────────────────
Total properties:   Jr = 2122 kg⋅m² (top)
                    Jb = 472 kg⋅m² (bottom)
                    k = 861.5 N⋅m/rad

Translating Simulation to Operational Limits

Simulation predicted: Stick-slip occurs at Wb = 42000 N and cb = 100 N⋅m⋅s/rad

Operational translation (from the thesis):

OPERATING ENVELOPE:
─────────────────────────────────────────────────
WOB Range           Safe RPM Windows    Drilling Mode
─────────────────────────────────────────────────
< 41000 N          All speeds OK        Stable (no damping control needed)
41000-41500 N      80-90 RPM            Slow, conservative
41500-42000 N      50-70 RPM            Critical zone (risky)
> 42000 N          AVOID                Stick-slip likely
─────────────────────────────────────────────────

Parallel damping strategy:
• Increase mud viscosity → raises cb → improves damping
• But keep cb < 150 N⋅m⋅s/rad (avoid permanent stalling)
• Optimal range: cb = 50-100 N⋅m⋅s/rad

Theoretical Operational Guidelines (From Simulation)

These are simulation-based predictions, not field-proven procedures. Based on the MSOG thesis model, the following parameters were identified as critical:

1. Critical parameter threshold (from simulation):

   Spring stiffness: k = (shear modulus × geometry) / length
   Prediction: Reducing k below ~500 N⋅m/rad suppresses stick-slip in model
   (This would correspond to longer drill strings)
   

2. Simulated operational window:

   Weight-on-Bit sensitivity: System shows instability transition around 42,000 N
   (2.4% change in WOB caused failure in simulation)
   
   Damping sensitivity: Optimal range in simulation was 50-100 N⋅m⋅s/rad
   (Below 50: unstable; Above 150: permanent stalling)
   
   RPM/Torque relationship: Higher torque alone didn't solve stick-slip
   (Suggests coupled parameter optimization needed, not single-variable control)
   

3. Simulated indicators (in MATLAB, not real drilling):

   • Torque oscillating between friction peaks (3300-5227 Nm) = stick-slip signature
   • Bit velocity discontinuously jumping 0 → 4.5 rad/s = phase transition occurring
   • Stick duration increasing with certain parameter combinations = approaching stall

Important caveat: These predictions come from a two-DOF model with simplified friction representation. Real drilling involves additional complexities (formation variability, temperature effects, lateral coupling, nonlinear bit-rock interactions) that the model doesn’t capture. Actual operational guidelines would require field validation.

Simulation Scope and Limitations

Important note: This analysis is entirely simulation-based. The thesis does not include experimental validation on a physical test rig or field testing. The results represent theoretical predictions from the MATLAB model.

What was simulated:

• Two-DOF torsional system with nonlinear friction model
• Stick-slip behavior under various parameter combinations
• AST tool performance under the same simulated conditions

What was not validated:

• Actual field drilling data comparison
• Test rig experiments
• Real-world AST performance
• Formation-specific friction behavior

The model makes several simplifying assumptions (linear elasticity, constant material properties, neglected lateral/axial coupling) that may not hold in actual drilling. Real field behavior would include:

• Variable formation properties with depth
• Temperature effects on material stiffness
• Fluid-structure interaction effects
• Nonlinear bit-rock interactions beyond the friction model
• Environmental uncertainties

Significance: The thesis demonstrates that mathematical simulation can predict stick-slip behavior and explore mitigation strategies computationally. These results provide a foundation for understanding torsional vibration mechanisms, but real-world application would require field validation before operational deployment.

Key Takeaways

1. Simulation can reveal mechanisms: Mathematical models can demonstrate how stick-slip oscillations develop and which parameters matter most. The MSOG model shows friction discontinuities drive the instability, not just high torque.

2. Parameter sensitivity is extreme: A 2.4% change in WOB shifted the system from stable to complete stalling. A single friction coefficient difference changed outcomes dramatically. This suggests real drilling is delicately balanced.

3. The friction model is critical: The exponential decay of friction with velocity (equation 3.12) is the key to understanding oscillations. Without it, the system wouldn’t exhibit stick-slip. This suggests friction characterization in real drilling is essential.

4. Damping has nonlinear effects: In simulation, damping doesn’t follow a simple “more is better” rule. Too little causes oscillations; too much causes stalling. An optimal window exists (50-100 N⋅m⋅s/rad in this model).

5. Spring stiffness matters more than expected: Reducing torsional stiffness (by lengthening the string) suppressed stick-slip in simulation. This counterintuitive result deserves field investigation.

6. AST concept is theoretically sound: The spring-based Anti-Stick-Slip tool prevented oscillations in the model by breaking the friction feedback loop. Whether this translates to field benefit requires real-world testing.

7. Model assumptions limit applicability: The two-DOF model simplifies drilling reality (constant properties, linear elasticity, no lateral coupling, no temperature effects). Real drilling complexity is greater than the model captures.

8. MATLAB implementation is transparent: The actual code (Appendix B) is straightforward enough for engineering teams to understand, modify, and validate. This democratizes access to vibration analysis—you don’t need proprietary software.