Latest / Mechanical Engineering Made Simple / (14) The Non-Linear Truth Shock and Vibration Engineering Fundamentals.

Transcript

(14) The Non-Linear Truth Shock and Vibration Engineering Fundamentals. · Mechanical Engineering Made Simple

• 0:00
  Welcome back to Mechanical Engineering Made Simple.
• 0:02
  Today we are pushing past those idealized classrooms, you know,
• 0:07
  where everything is clean, light, and totally friction
• 0:10
  free. We are taking a deep dive into
• 0:13
  the well, the messy reality of mechanical engineering,
• 0:16
  nonlinearity, instability, and how we actually control
• 0:20
  vibration when the equations frankly refuse to behave.
• 0:23
  It's so true. Historically, mechanical
• 0:25
  analysis really relied heavily on linearization, which is, you
• 0:30
  know, essentially an indispensable shortcut.
• 0:32
  We take these complex physical laws, maybe involving fluid
• 0:35
  forces or large displacement geometry, and we simplify them,
• 0:40
  make them into linear equations that are, well, easy to solve.
• 0:44
  This gives us a decent first look and analysis of an
• 0:46
  idealized system, and it's often close enough provide the motions
• 0:50
  stay really small. But engineering systems, they
• 0:53
  rarely stay small or tidy, do they?
• 0:55
  The moment you introduce significant vibration amplitudes
• 0:57
  or specialized materials that don't quite follow Hook's law or
• 1:01
  think about the critical clearances in high speed
• 1:03
  machinery, those clean linear assumptions, they just
• 1:06
  breakdown. So our mission today is to
• 1:08
  unpack what happens when those assumptions fail.
• 1:11
  We'll look at the really ingenious ways we fight back
• 1:13
  with vibration control systems, but also the catastrophic
• 1:17
  failures caused by things like self excited motion.
• 1:21
  OK, let's unpack this. Let's start with the fundamental
• 1:23
  problem. Why does linearity fail?
• 1:27
  Where does it fall apart? Well, when we're forced to
• 1:29
  incorporate the full physics of a system, we step into the realm
• 1:33
  of non linear analysis. And what's fascinating here
• 1:37
  truly, is that non linearity doesn't just tweak the linear
• 1:40
  answer. It often reveals totally
• 1:42
  unsuspected phenomena, like brand new behaviors that linear
• 1:45
  theory simply couldn't predict, not even hint at.
• 1:47
  I guess the most accessible example is the simple pendulum,
• 1:50
  right? If you remember from intro
• 1:51
  physics we always assume the angle Theta is small, so we can
• 1:55
  approximate sin Theta as just Theta.
• 1:57
  Exactly. That's the linear equation, and
• 1:59
  it tells you the period is constant no matter how high you
• 2:01
  swing it, which feels intuitive but isn't quite right.
• 2:05
  But if you analyze the full exact equation, the word
• 2:09
  involving sin Theta, you find something different.
• 2:11
  The period of oscillation actually increases as the
• 2:14
  amplitude increases. That relationship period versus
• 2:18
  amplitude. That's a purely non linear
• 2:20
  effect, totally missed by the linear model, but pretty crucial
• 2:24
  if you're building, say, a highly accurate clock.
• 2:27
  And it's not just motion geometry, right?
• 2:28
  It's materials to hooks. Law assumes that perfect
• 2:32
  straight line relationship force equals stiffness times
• 2:36
  displacement. But real springs or complex
• 2:39
  structures, they deviate. Yeah, they do.
• 2:42
  And these deviations, we characterize them based on how
• 2:45
  that restoring force changes relative to the displacement.
• 2:48
  We tend to classify them into two main types of stiffness
• 2:50
  characteristics. OK, what's the first one?
• 2:52
  The first is the hardening characteristic.
• 2:54
  This happens when the restoring force ramps up more rapidly than
• 2:56
  the displacement. Think about a coil spring where
• 2:59
  the coils actually start closing up, touching each other.
• 3:02
  The force needed to compress it further just skyrockets, right?
• 3:05
  Mathematically, we often model this by adding a cubic term, an
• 3:09
  X ^3 term with a positive sign. It makes the stiffness curve
• 3:14
  sweep upward. OK, hardening makes sense.
• 3:16
  So the opposite. Conversely, yes, we have the
• 3:19
  softening characteristic here. The restoring force increases
• 3:22
  less rapidly than displacement as the amplitude grows.
• 3:25
  We actually just saw this with the pendulum example.
• 3:28
  The larger the swing, the less effective that restoring gravity
• 3:30
  component becomes, relatively speaking.
• 3:33
  And this is modeled using an X ^3 term too, but this time with
• 3:36
  a negative sign. It makes the curve sweep
• 3:38
  downward. That X ^3 term is particularly
• 3:41
  useful because it maintains the physical symmetry.
• 3:44
  Yeah, the restoring force acts the same regardless of the
• 3:48
  direction displacement. And crucially, I think this is
• 3:50
  important. Even if all your individual
• 3:52
  components are perfectly linear, the sister itself can still
• 3:55
  become non linear just through geometry or physical
• 3:58
  constraints. The source material highlights
• 4:00
  something called snubber action. Like imagine a mass oscillating
• 4:03
  linearly until it hits a rigid stop.
• 4:06
  A hard limit. Yeah, that collision instantly
• 4:09
  changes the stiffness profile. It leads to what's called a
• 4:12
  piecewise linear characteristic. So the overall system is now
• 4:16
  behaving non linearly even though the internal components
• 4:20
  are all, you know, clean linear springs.
• 4:22
  Exactly the system behavior emerges from the interaction.
• 4:26
  So OK, if we accept the real world is messy, non linear, the
• 4:30
  engineer's first job is often control.
• 4:32
  We have systems constantly getting hit by some persistent
• 4:36
  external harmonic force, what we call forced vibration.
• 4:39
  How do we fix it? The approach seems to be
• 4:42
  introducing auxiliary mass systems.
• 4:44
  That's right. And there are two main
• 4:45
  strategies here really, depending on whether that
• 4:47
  excitation frequency is constant or if it might vary.
• 4:50
  The first solution is the dynamic absorber.
• 4:53
  This is used only when the excitation frequency is
• 4:55
  precisely known and doesn't change.
• 4:57
  It's basically just an auxiliary mass attached via a spring, and
• 5:00
  it's tuned very precisely exactly to that excitation
• 5:04
  frequency. And here's the the aha moment,
• 5:06
  the mechanism that makes it so clever.
• 5:08
  When that absorber is perfectly tuned, it effectively creates a
• 5:12
  point of no motion, a node right where it attaches to the main
• 5:14
  system. It forces it to be still.
• 5:16
  Exactly. It works because the force
• 5:19
  generated by that absorber mass oscillating, which is F equals
• 5:23
  Omega squared on ought the mass times frequency squared times
• 5:26
  absorber amplitude. That force exactly cancels out
• 5:30
  the incoming external force, so the primary mass it just stays
• 5:33
  perfectly still. It's quite elegant.
• 5:35
  It is elegant, but the limitation is, well, severe.
• 5:38
  As soon as that incoming frequency shifts even a tiny bit
• 5:41
  of fraction of a percent, the absorber stops absorbing.
• 5:45
  Worse, it can actually bring the entire combined system into a
• 5:48
  dangerous new state of residence.
• 5:51
  So it turns the fix into a potential disaster.
• 5:53
  Pretty much if the conditions aren't perfect.
• 5:55
  OK, here's where it gets really interesting because that brings
• 5:59
  us to the more robust solution, right?
• 6:01
  The auxiliary mass damper or sometimes called a damped
• 6:05
  absorber. This system adds A viscous
• 6:08
  damper element between the auxiliary mass and the primary
• 6:11
  structure that makes it effective even if the excitation
• 6:15
  frequency wobbles a bit. That's the key improvement, and
• 6:18
  the optimal design for this system relies on a really clever
• 6:21
  insight you get from plotting the frequency response curve.
• 6:24
  When you vary the damping parameter Zeta in a damped
• 6:27
  absorber system, the response curves for different damping
• 6:31
  values all intersect at 2 specific points.
• 6:33
  We call these locations fixed points, usually labeled A&B.
• 6:36
  Fixed points and why are they so crucial?
• 6:39
  Because the vibration amplitude at those two points is
• 6:41
  completely independent of how much damping you apply.
• 6:44
  Think about it. Normally adding damping lowers
• 6:46
  the vibration curve everywhere, right?
• 6:47
  Reduces the peaks. But at points A&B the amplitude
• 6:51
  just stays put. Doesn't matter how much you
• 6:53
  tweak the damper. So the engineer's goal then
• 6:55
  becomes setting the optimal tuning ratio alpha opt so that
• 6:59
  the amplitude at point A is exactly equal to the amplitude
• 7:03
  at point B. You level them out.
• 7:05
  Exactly because by leveling the response at these two crucial
• 7:10
  damping independent points, you essentially guarantee that the
• 7:13
  maximum peak response the system will ever hit across a range of
• 7:16
  operating frequencies is minimized.
• 7:19
  You're actively designing for the flattest, safest possible
• 7:22
  response curve. That optimization, that's how
• 7:26
  heavy industry keeps things stable then.
• 7:28
  And this isn't just abstract theory is it?
• 7:30
  This is what powers real world things like ship anti roll
• 7:33
  systems. They use tanks of fluid pumped
• 7:34
  around or sometimes massive gyroscopes.
• 7:36
  Right, acting as auxiliary masses.
• 7:38
  Acting as auxiliary masses to counteract that persistent
• 7:41
  rolling frequency the ship experiences from waves.
• 7:44
  That's forced vibration control enact.
• 7:46
  Precisely. So OK, if we connect this back
• 7:50
  to the bigger picture, we've been talking about forced
• 7:53
  vibration, something driven from the outside.
• 7:55
  Now we need to shift gears to what is arguably the ultimate
• 7:58
  danger in mechanical systems, self excited vibration.
• 8:02
  This is fundamentally different and often far, far more
• 8:05
  destructive. How is it different?
• 8:07
  What's the core idea? In a self excited system, the
• 8:10
  alternating force that sustains the motion is actually generated
• 8:14
  or controlled by the motion itself.
• 8:16
  If the system stops moving, the force just disappears entirely.
• 8:20
  OK, so the motion feeds the force that causes the motion a
• 8:24
  feedback loop. Exactly.
• 8:26
  It's why the source material explains it mathematically as
• 8:28
  being analogous to negative damping in a linear system.
• 8:31
  The solution involves an exponentially increasing
• 8:34
  amplitude. It heads straight towards
• 8:35
  instability and likely system failure.
• 8:38
  And most critical place that shows up is in high speed
• 8:41
  rotating machinery. Yes, leading to whirling and
• 8:44
  whipping of shafts. Think jet engines, turbines,
• 8:47
  pumps. This instability is driven by a
• 8:50
  tangential force. It acts normal or perpendicular
• 8:54
  to the radial deflection of the shaft.
• 8:56
  And crucially, this force is proportional to that deflection.
• 9:00
  So the more it deflects, the stronger the driving force.
• 9:03
  It acts like a little engine, continuously feeding energy into
• 9:07
  the whirling motion. And it overcomes the natural
• 9:09
  damping in the system that. Overwhelms any external damping,
• 9:12
  yeah, causing the shaft to spin out of control, basically.
• 9:14
  Now what's vital for you know for you listening to GRASP is
• 9:18
  the diagnostic feature. How do you tell this apart from
• 9:22
  just simple imbalance which is also common?
• 9:24
  Good point. In self excited world, the
• 9:26
  whirling motion happens at the shafts own natural or critical
• 9:29
  frequency. And here's the kicker, that
• 9:32
  whirl frequency is entirely independent of the shafts
• 9:34
  rotational speed, the RPM. OK, so if the machine spins
• 9:38
  faster or slower, the whirl frequency stays locked onto the
• 9:42
  system's natural frequency. Exactly.
• 9:44
  Whereas vibration due to imbalance is always synchronous.
• 9:47
  It happens at the rotation frequency.
• 9:49
  And here's the profound danger. Simple synchronous force
• 9:53
  vibration from imbalance. It's bad, but somewhat
• 9:56
  predictable. The stress on a given point in
• 9:59
  the shaft is constant. As it rotates.
• 10:01
  Self excited whirl induces intense alternating stresses as
• 10:06
  the shaft whirls. The metal fibers on the surface
• 10:08
  are constantly being bent and straight and flexed and relaxed
• 10:12
  at the whirl frequency. This leads to rapid catastrophic
• 10:15
  fatigue failure. That stress reversal is far more
• 10:18
  damaging than the constant stress from synchronous
• 10:21
  unbalance. That makes sense, rapid cycling
• 10:23
  breaks things faster. So let's quickly run down the
• 10:25
  common culprits, the mechanisms behind this self excited
• 10:28
  instability in machines. OK, first there's hysteretic
• 10:31
  whirl. This is caused by internal
• 10:33
  friction material hysteresis, like within the shaft material
• 10:37
  itself or maybe in shrink fits or joints.
• 10:40
  Internal damping, but acting the wrong way here.
• 10:42
  Right. This internal friction causes a
• 10:44
  tiny phase shift between the bending stress and the strain as
• 10:48
  the shaft flexes, and this results in a small residual
• 10:52
  tangential force that drives forward whirl, meaning whirling
• 10:56
  in the same direction as the shaft rotation.
• 10:58
  Crucially, this type usually only kicks in once the shaft
• 11:01
  speed gets above its first critical speed.
• 11:03
  Correct. Needs enough speed to overcome
• 11:05
  the stabilizing effects. OK, second, highly common,
• 11:07
  especially in power generation. That would be fluid bearing whip
• 11:11
  or sometimes oil whip. Here the viscous fluid.
• 11:14
  The oil circulating in the bearing clearance creates
• 11:17
  asymmetric pressure forces as the shaft moves off center.
• 11:20
  These generate a powerful tangential force because the
• 11:24
  fluid usually circulates at roughly half the shaft speed.
• 11:27
  The famous half frequency. Whirl exactly.
• 11:29
  This instability typically requires the shaft RPM to be
• 11:32
  greater than or equal to twice its critical speed before it
• 11:35
  really kicks in. It also causes forward.
• 11:37
  Roll got it. And finally, the really scary
• 11:39
  one. Dry friction whip.
• 11:42
  Yeah, this is often called rubbing.
• 11:43
  It's a terrifying event that happens when the rotating shaft
• 11:46
  actually makes physical contact with a stationary part, like a
• 11:49
  seal or maybe the stator casing. The resulting Coulomb friction
• 11:54
  generates a tangential force. It's highly sensitive to the
• 11:58
  contact conditions, but it often indexes backward whirl.
• 12:01
  Backward whirl so whirling against the direction of
• 12:05
  rotation. Yes, you can almost picture the
• 12:07
  shaft trying to roll or climb along the static boundary it's
• 12:10
  rubbing against very destructive very quickly.
• 12:13
  OK, we've seen control methods. We've seen these catastrophic
• 12:16
  self excited failures. Let's wrap up by looking briefly
• 12:19
  at what happens when you bring nonlinearity back into a force
• 12:23
  system, even one that might seem stable initially.
• 12:26
  You get these strange multi valued responses and ultimately
• 12:30
  chaotic behavior, right? One of the absolute hallmarks,
• 12:33
  especially in systems that have that hardening or softening
• 12:35
  stiffness we talked about, is the jump phenomenon.
• 12:37
  Remember those stiffness curves that sweep up or down when you
• 12:40
  plot the force vibration response?
• 12:42
  For these nonlinear systems, the resonance peaks aren't symmetric
• 12:45
  anymore. That gets significantly swept
• 12:47
  over to the right for hardening systems and to the left for
• 12:50
  softening ones. So what happens if you're
• 12:52
  operating near that bent over peak?
• 12:54
  Well, let's say you slowly increase the frequency of the
• 12:57
  external force on a hardening system.
• 13:00
  The vibration amplitude rises along the curve, but then
• 13:04
  instead of smoothly following the curve over the top, it hits
• 13:07
  a point where the curve becomes vertical, a point of vertical
• 13:10
  tangency. At that exact point, the system
• 13:13
  physically cannot sustain that high amplitude state and it will
• 13:17
  suddenly instantaneously jump down to a much lower stable
• 13:21
  response level on a different branch of the curve A.
• 13:24
  Sudden drop in vibration. Exactly.
• 13:27
  And conversely, if you then slowly decrease the frequency,
• 13:30
  the system stays on that low amplitude branch until it hits a
• 13:33
  different point of vertical tangency, forcing it to suddenly
• 13:36
  jump up back to the high amplitude branch.
• 13:39
  Wow, so there's a whole range of frequencies where you can be in
• 13:42
  either a high vibration state or a low vibration state, depending
• 13:45
  on whether you got there by increasing or decreasing
• 13:48
  frequency. Precisely, it reveals this
• 13:50
  critical aspect. In that frequency range, there
• 13:53
  are often 3 mathematically possible solutions for the
• 13:56
  amplitude. But that intermediate solution,
• 13:59
  the one connecting the jump points, it's fundamentally
• 14:02
  unstable. The system cannot physically
• 14:04
  operate there. Any tiny disturbance forces it
• 14:07
  to jump to one of the stable high or low states.
• 14:10
  That has huge implications for a design and testing, I imagine.
• 14:13
  And the ultimate breakdown of predictability, sort of the
• 14:15
  final frontier mentioned in the source material, is chaotic
• 14:19
  dynamics. Yeah, chaos.
• 14:22
  Chaotic vibrations are characterized by waveforms that
• 14:24
  look, well, irregular and non periodic.
• 14:26
  The pattern never actually repeats itself, but the defining
• 14:30
  feature mathematically is extreme sensitivity to initial
• 14:33
  conditions if you start a chaotic system twice with
• 14:36
  starting points that are almost identically the same.
• 14:38
  Infinitesimally different. The resulting motions, the
• 14:41
  solutions, will diverge exponentially over time.
• 14:44
  They become completely different.
• 14:46
  This means they are inherently unpredictable in the long term.
• 14:48
  Long term forecasting or control becomes essentially impossible.
• 14:52
  Sounds like just random noise then.
• 14:54
  It looks like it sometimes, but there's often hidden structure
• 14:58
  and engineers have found ways to visualize this.
• 15:01
  The main tool is the Poincare section.
• 15:03
  It's like using a strobe light on the motion.
• 15:05
  You plot only the displacement versus the velocity, but you
• 15:09
  only take a snapshot a single point exactly once per cycle of
• 15:13
  the external forcing function. So you're sampling the state of
• 15:16
  the system at regular intervals based on the driving force.
• 15:19
  Exactly. And instead of getting a
• 15:20
  completely random smear of data points, which is what you'd
• 15:23
  expect from truly random noise, the Poincare section of a
• 15:27
  chaotic system often reveals the striking, intricate, but
• 15:30
  definitely structured patterns. These complex bounded structures
• 15:34
  are called attractors, sometimes strange attractors.
• 15:37
  They give us insight into the boundaries and the nature of the
• 15:40
  chaotic motion. Even if we can't predict the
• 15:42
  exact path, it helps us understand the limits of the
• 15:45
  chaos. That journey then, from the
• 15:48
  simple linear ideal of a pendulum, through precise
• 15:52
  vibration controls like dynamic absorbers and damped absorbers,
• 15:56
  all the way through self destruction from whirling and
• 15:58
  whipping and landing and outright chaos, It really
• 16:01
  reveals just how much hidden complexity lies beneath the
• 16:04
  surface of seemingly simple rotating machines.
• 16:07
  Absolutely. This raises an important
• 16:09
  question. We often think about mechanical
• 16:10
  stability purely in terms of damping, right?
• 16:13
  Does it dissipate energy? But the stability of even a
• 16:15
  simple periodic motion in a really complex nonlinear system,
• 16:19
  Yeah. It might be governed by much
• 16:20
  deeper concepts, like Matthew's equation, for example.
• 16:23
  That equation statistically characterizes whether small
• 16:25
  deviations away from a known periodic motion will grow
• 16:28
  exponentially or decay back. So the final thought for you,
• 16:31
  the listener, is maybe this. You should think about the
• 16:33
  stability of any mechanical design, not just in terms of
• 16:36
  whether it dampens external forces effectively, but ask if
• 16:41
  the system deviates just slightly from its expected
• 16:43
  motion, does that deviation push it towards exponential
• 16:47
  instability, or maybe even into chaotic behavior?
• 16:49
  Understanding that inherent fragility, that potential for
• 16:52
  small errors to cascade, that's maybe the true deep dive take
• 16:56
  away here. Absolutely.
• 16:57
  The engineering challenge is always recognizing that while
• 17:00
  our linear models get the job done maybe 99% of the time,
• 17:04
  yeah, the catastrophic failure often lies hidden in those
• 17:07
  nonlinear terms we initially ignored.
• 17:09
  We really appreciate you joining us for this deep dive.
• 17:12
  We'll see you next time on Mechanical Engineering Made
• 17:14
  Simple.