ME451 Kinematics and Dynamics of Machine Systems Newton-Raphson Method 4.5 Singular Configurations of Mechanisms 3.7 March 5, 2009 © Dan Negrut, 2009 ME451, UW-Madison

Before we get started… Last Time Discussed Newton’s method for solving nonlinear systems Needed when solving  (q,t)=0 to determine the generalized positions q at each time t Newton method requires two things: A starting point q (0) from where the solution starts being searched for An iterative process in which the approximation of the solution is gradually improved: 2 Announcements Exam coming up next Tu (Mar. 10) Take-home component posted online, due Friday, Mar. 13 (11:59 PM) You can bring to exam one sheet of paper (both sides) with any equations/information that you find useful

Take-Home Exam Component This component of the exam is worth 40% of the exam score More challenging than it seems, start early, particularly if you are not familiar with MATLAB Something other than MATLAB might be acceptable (open to alternatives) You will implement, using MATLAB code discussed on Tu Kinematics Analysis + Design Sensitivity Study 3

Middle of the semester survey Anonymously, please print out on a sheet of paper and bring in on Th (one week from now): Two things that you don’t like about the class and I should avoid doing in the future Any other comments that you might have I’ll get back with you to summarize your comments during next class 4

Newton’s Method: Closing Remarks Can ever things go wrong with Newton’s method? Yes, there are at least three instances: 1. Most commonly, the starting point is not close to the solution that you try to find and the iterative algorithm diverges (goes to infinity) 2. Since a nonlinear system can have multiple solutions, the Newton algorithm finds a solution that is not the one sought (happens if you don’t choose the starting point right) 3. The speed of convergence of the algorithm is not good (happens if the Jacobian is close to being singular (zero determinant) at the root, not that common) 5

Newton’s Method: Closing Remarks What can you do address these issues? You cannot do anything about 3 above, but can fix 1 and 2 provided you choose your starting point carefully Newton’s method converges very fast (quadratically) if started close enough to the solution To help Newton’s method in Position Analysis, you can take the starting point of the algorithm at time t k to be the value of q from t k-1 (that is, the very previous configuration of the mechanism) 6

Position Analysis: Final Form 7 % Driver for Position Analysis % Works for kinematic analysis of any mechanism % User needs to implement three subroutines: % provideInitialGuess % getPhi % getJacobian % general settings normCorrection = 1.0; tolerance = 1.E-8; timeStep = 0.05; timeEnd = 4; timePoints = 0:timeStep:timeEnd; results = zeros(3,length(timePoints)); crntOutputPoint = 0; % start the solution loop q = provideInitialGuess(); for time = 0:timeStep:timeEnd crntOutputPoint = crntOutputPoint + 1; while normCorrection>tolerance, residual = getPhi(q, time); jacobian = getJacobian(q); correction = jacobian\residual; q = q - correction; normCorrection = norm(correction); end results(:, crntOutputPoint) = q; normCorrection = 1; end function phi = getPhi(q, t) % computes the violation in constraints % Input : current position q, and current time t % Output: returns constraint violations dummy1 = q(1) *sin(q(3)); dummy2 = q(2) *cos(q(3)); dummy3 = q(1)*q(1) + q(2)*q(2) - (t/10+0.4)*(t/10+0.4); phi = [ dummy1 ; dummy2 ; dummy3 ]; function jacobian = getJacobian(q) % computes the Jacobian of the constraints % Input : current position q % Output: Jacobian matrix J jacobian = [ *cos(q(3)) ; *sin(q(3)); 2*q(1) 2*q(2) 0 ]; function q_zero = provideInitialGuess() % purpose of function is to provide an initial % guess to start the Newton algorithm at % time=0 q_zero = [0.2 ; 0.2 ; pi/4];

END: Newton Method BEGIN: Singular Configurations of Mechanisms Section 3.7 8

Singular Configurations What are “singular configurations”? Abnormal situations that should be avoided since they indicate either a malfunction of the mechanism (poor design), or a bad model associated with an otherwise well designed mechanism Singular configurations come in two flavors: Physical Singularities (PS): reflect bad design decisions Modeling Singularities (MS): reflect bad modeling decisions Singular configurations do not represent the norm, but you must be aware of their existence A PS is particularly bad and can lead to dangerous situations 9

Singular Configurations In a singular configuration, one of three things can happen: PS1: Your mechanism locks-up PS2: Your mechanism hits a bifurcation MS1: Your mechanism has redundant constraints The important question: How can we characterize a singular configuration in a formal way such that we are able to diagnose it? Look at two examples to see what happens in a singular configuration 10

Mechanism Lock-Up: PS1 (Example 3.7.5, draws on 3.1.2) Investigate what happens to this mechanism when length l= Can you ever get in trouble? Yes, check what happens when t=2 Mechanism hits a lock-up configuration When t=2:

Definition of lock-up configuration: The mechanism cannot proceed anymore Symptoms of “lock-up”: Jacobian in that configuration is singular The rank of the velocity augmented constraint Jacobian is higher than the rank of the constraint Jacobian The velocities and accelerations assume huge values (in fact, going to infinity)  That is, you’re sure not to miss it… 12 Mechanism Lock-Up Velocity augmented constraint Jacobian

Mechanism Lock-Up (Cntd.) Investigate rank of augmented Jacobian Carry out velocity analysis Carry out acceleration analysis 13 time = 1.85 vel = [ ] time = 1.90 vel = [ ] time = 1.95 vel = [ ] time = 2.00 vel = 1.0e+007*[ ] Mechanism moves faster than speed of light… time = 1.80 acc = [ ] time = 1.85 acc = [ ] time = 1.90 acc = [ ] time = 1.95 acc = [ ] time = 2.00 acc = 1.0e+022*[ ]

Bifurcation: PS2 (Example 3.7.5, draws on 3.1.2) Investigate what happens to this mechanism when length l=1 14 Can you ever get in trouble? Yes, check what happens when t=6 Mechanism hits a bifurcation When t=6:

Definition of bifurcation configuration: The mechanism can proceed in more than one way Symptoms of “bifurcation”: Jacobian in that configuration is singular The rank of the velocity and acceleration augmented constraint Jacobians is equal to the rank of the constraint Jacobian The velocities and accelerations do not assume huge values  That’s why it’s tough to spot a bifurcation (unlike a lock-up), often times you cruise through it without knowing it… 15 Bifurcation (Cntd.) Acceleration augmented constraint Jacobian

Bifurcation, Scenario 1: Time Step is 0.06 [s] Investigate rank of augmented Jacobians Carry out velocity analysis Carry out acceleration analysis 16 time = 5.80 vel = [ ] time = 5.86 vel = [ ] time = 5.92 vel = [ ] time = 5.98 vel = [ ] time = 6.04 vel = [ ] time = 6.10 vel = [ ] time = 6.16 vel = [ ] time = 5.80 acc = [ ] time = 5.86 acc = [ ] time = 5.92 acc = [ ] time = 5.98 acc = 1.0e-003 *[ ] time = 6.04 acc = 1.0e-012 *[ ] time = 6.10 acc = 1.0e-012 *[ ] time = 6.16 acc = 1.0e-013 *[ ] NOTE: Stepped over bifurcation configuration and hardly noticed Stepping over bifurcation... Bifurcation Time: T=6

Carry out velocity analysis Carry out acceleration analysis 17 time = 5.85 vel = [ ] time = 5.90 vel = [ ] time = 5.95 vel = [ ] time = 6.00 vel = [ NaN NaN -Inf] Warning: Matrix is singular to working precision. > In function bifurcation at line 14 time = 6.05 vel = [ ] time = 6.10 vel = [ ] time = 6.15 vel = [ ] time = 5.85 acc = [ ] time = 5.90 acc = [ ] time = 5.95 acc = [ ] time = 6.00 acc = [NaN NaN NaN] Warning: Matrix is singular to working precision. > In function bifurcation at line 19 time = acc = 1.0e-011 *[ ] time = acc = 1.0e-012 *[ ] time = acc = 1.0e-012 *[ ] time = acc = 1.0e-013 *[ ] Bifurcation, Scenario 2: Time Step is 0.05 [s] Bifurcation Time: T=6 NOTE: On previous slide we were “lucky”, we chose a step-size that happen to hit right on the bifurcation

Singular Configurations In the end, what is the pattern that emerges? The punch line: The only case when you run into problems is when the constraint Jacobian becomes singular: Otherwise, the Implicit Function Theorem (IFT) gives you the answer: If the constraint Jacobian is nonsingular, IFT says that you cannot be in a singular configuration. And that’s that. 18

Singularities: Closing Remarks Remember that you seldom see singularities To summarize, if the constraint Jacobian is singular, You can be in a lock-up configuration (you won’t miss this, PS1) You might face a bifurcation situation (very hard to spot, PS2) You might have redundant constraints (we didn’t say anything about this, MS1) Singularity analysis is a tough topic. Textbook gives a broader perspective, although not necessarily deeper More thorough treatment in ME751 (particularly for redundant constraints, MS1) Some small mistakes in text 19

Perform Position Analysis for the mechanism below Example of Mechanism Analysis NOTE: the result of this analysis is you being able to say, given the two motions, where each body is, and how it is oriented at each moment in time 20

Strategies for Kinematics Analysis You can embrace one of two strategies to carry out Kinematics Analysis They are different based on the number of generalized coordinates used to carry out the analysis Strategy 1: use a reduced set of generalized coordinates Strategy 2: use Cartesian generalized coordinates 21

The Reduced Set Strategy Advantages Few generalized coordinates lead to few kinematic constraints (that is, number of equations in  (q,t)=0 is small) For small mechanisms (3-4 bodies), easy to solve with pencil and paper Disadvantages This strategy is not general (systematic), but rather is applied on a case by case situation You are back to a ME240 situation, where each problem comes with its own solution Not trivial to use for large systems, especially when you are dealing with 3D mechanisms 22 Reduced set of generalized coordinates: (Example 3.5.5)

The Cartesian Set Strategy Advantages This strategy is general (systematic), it always works (used in ADAMS as well) It is rather mechanical in nature, there is no thinking involved, it’s just simply following a recipe It relies on a very limited number of building blocks (provided in book) to implement a systematic approach that allows for analysis of any mechanism no matter how large it is Disadvantages Larger set of generalized coordinates leads to larger number of kinematic constraints (that is, number of equations in  (q,t)=0 is large) For simple systems it’s an overkill (the mosquito and the cannonball) 23 Cartesian generalized coordinates: (Example 3.5.5)

The Cartesian Set Strategy: Example 24 For the wrecker boom mechanism, use Cartesian coordinates and carry out the steps required by Kinematic Analysis Two motions prescribed to control the motion of the boom: