In acoustic tomography, a convenient parameter to measure is the transit time (time-of-flight) from source to receiver. Hence, in application to time-of-flight (TOF) tomography, it is of interest to reconstruct the tomographic image which is a distribution of slowness (the inverse of acoustic velocity) throughout the image.

In the present investigation, TOF tomography has been used to reconstruct slowness in a simulated composite domain consisting of an isotropic base material containing an insert of a resin rich zone. Algebraic reconstruction techniques (ART) have been employed for computing the slowness distribution (in the domain) in an iterative manner from TOF projection data. The main objective of this work is to,

• observe the variation in the response at the receiver end when the concerned ray passes (or bypasses) through the insert region and,
• propose certain guidelines regarding collection of the TOF data from the simulated response.

For collecting the projection (TOF) data, simulations using a two-dimensional finite element (FE) model have been performed. The FE formulation simulates pulsed excitation at specific nodes (input node) on the boundary of the domain in the direction of the concerned "view" for a finite pulse duration. The fE analysis solves for displacement components at each node at every time step. In this study, the displacement response is recorded at the boundary node situated on the corresponding ray (output node). The displacement field for the output node is then plotted against time and the TOF of the acoustic wave from input to output node is evaluated.

The main observations found in this work can be summarised as follows.

1. A guide line is proposed for collection of ultrasonic TOF data for actual experiments. Any response with close peaks indicates essentially the existence of an insert in the vicinity of the ray. In such cases, TOF should be taken as the initiation time corresponding to the higher peak portion of the wedge. Several cases have been checked and verified with results computed analytically.

2. It is also seen that, as expected theoretically, 4-view reconstructions are quantitatively much better than the 2- view result and it is expected that further enhancement in the reconstructed image is possible by increasing number of views and choosing appropriate relaxation parameters in the ART algorithm.