Software project on fracture dynamics of wood and artificial neural network model for prediction of fracture toughness

Abstract

This study focussed on fracture and dynamics in wood using high speed imaging equipment, general and various types of imaging software other than the general equipment such as testing machine and related software. The C++ programming language helped to analyse and calculate data. Recorded fracture processes at experiments were highly valuable as they carry very important dynamic data rather than animation. This study clearly displayed the importance of the crack closure and extensions at fracture, which provided quantitative and qualitative data with the aid of software. Further data obtained from this study on fracture dynamics that combines the physical fracture path and relevant load or stress along the fracture path during its fracturing time events seems to be more important. But the thesis limits the analysis up to quantitative values of physical fracture such as crack length and speed separately analysed with the fracture toughness for crack initiation.

Fracture toughness at crack initiation was modelled using an Artificial Neural Network software package. It was a hard task due to the complexity of the parameters related to fracture. However, a fine model was developed drawing zero weights to about 40% of the input parameters used. The model proved that the linked (uncracked) particles or molecules of wood are highly influenced on fracture and fracture toughness. It agrees with the Weibull's weakest link theory, but changes occur according to the loading configuration used. Therefore the local volume effect as the size effect is hypothesised to distance dependent from the crack plane and the loading plane. The study shows that the fracture effectiveness of the geometry factor of the fracturing member differs according to the loading configuration. In other words, the type of the geometry factor whether the volume or the length etc., is determined by the loading configuration used in the application. Therefore, a requirement of a proper definition or categorisation of loading configuration for fracture is raised at this point. This concept is effectively confirmed by the study in crack dynamics too as it shows the low volumetric effect at high rates of loading while it is high at low rates of loading.

The study has contributed considerable new work to its field. This includes a theoretically and practically sound method of deriving length and speed of individual crack of the bunch of cracks made at any event during the fracture process recorded by a high speed camera followed by a C++ programming module and a Dbase IV database. The data accuracy was very high due to this module and the use of several general and imaging software packages too. There are valuable data left behind that can be used for computer simulation type of studies on fracture. Approximation using Artificial Neural Network method on fracture toughness was also a new method for this study, which provided a very good model.