This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 4: Roundoff and Truncation Errors 4: Roundoff and Truncation Errors Sep 29, 2006 10:02 ECOR2606  Hassan & Holtz 2 Chapter Objectives To acquaint you with the major sources of errors involved in numerical methods. ● The distinction between accuracy and precision. ● The meaning of 'significant digits'. (n.i.t.) ● How to quantify error. ● How error estimates can be used to terminate iterative computations. ● Binary representation of numbers. (n.i.t.) ● Roundoff errors because digital computers have limited ability to represent numbers. ● Truncation errors because exact mathematical formulations are represented by approximations. ● Use of the Taylor series to estimate truncation errors. ● Forward, backward and centred finite difference approximations of first and second derivatives. Sep 29, 2006 10:02 ECOR2606  Hassan & Holtz 3 Sources of Errors ● Roundoff errors : inability of computers (and calculators) to exactly represent certain numerical quantities. ● Truncation errors : representation of mathematical formulations by approximations (e.g., finite difference approximation, see below). ● Formulation or modelling errors: the mathematical model is a poor or incorrect model of the physical world (e.g, g is constant). ● Data uncertainty: mechanical or physical properties may not be known with sufficient accuracy (e.g. drag coefficient). ● Blunders: errors in calculation, recording, etc. Only the first two of these will be considered here. EG: the following finitedifference approximation to a first derivative was used to compute v(t i+1 ) in the bungeejumper problem: dv dt ≈ v t = v t i 1 − v t i t i 1 − t i Sep 29, 2006 10:02 ECOR2606  Hassan & Holtz 4 4.1 Errors Accuracy – refers to how closely a computed or measured value agrees with the true value. Precision – refers to how closely repeated computed or measured values agree with each other. Figure 4.1: Inaccuracy ( bias ) – systematic deviation from the truth. Imprecision ( uncertainty ) – refers to the amount of “scatter”. Error – a term that refers to both. Sep 29, 2006 10:02 ECOR2606  Hassan & Holtz 5 Error Definitions True Error True Relative Error Approximate Relative Error True value = approximation error True relative error = true value − approximation true value a = approximate error approximation 100 % E t = true value − approximation t = true value − approximation true value 100 % a = present approximation − previous approximation present approximation 100 % 4.5 Sep 29, 2006 10:02 ECOR2606  Hassan & Holtz 6 Stopping Criteria Many numerical algorithms are iterations of the form: make initial estimate until error < maximum allowable error make an improved estimate ● do not know what the true answer is (obviously!)....
View
Full
Document
This note was uploaded on 11/13/2011 for the course CIVE 2*** taught by Professor  during the Spring '11 term at Carleton CA.
 Spring '11
 

Click to edit the document details