In my non-linear analysis code made in MATLAB 2013 by Newton Raphson method, the value is not converging at some point. But in other cases it worked fine.
Is it because of load increments,i have tried giving small load increments but its not working. Is there any way to adopt automatic load increments?
I tried using the formula (increment (i+1))=increment(i)*sqrt(no of desired iterations/no of iterations for previous increment)
Can anyone please suggest any way out?
Thank you for your replies sir...Max Demenkov and Yogendra Nath.
But what can I do now, just trial and error??
So for every type of problems I am going to do, Will I have to try with different initial load and load increments to get convergence...??
If f''(x)/f'(x) is large near the root then Newton's method will not converge, and so need to use a different method at those points. Not all methods work equally well for different equations. So need to choose the method depending on the equation.
See : https://www.math.ubc.ca/~anstee/math104/104newtonmethod.pdf
Try the bissection method which is generally more robust.
If you have initial interval, it's better to use bisection method, while it's slow. What you observe in your approach, is very probable in NR method, especially when your system has more than one root. For more fundamental view, you can see works on Julia set in fractal discussions.
- Badges
- Science topic