3D modelling, finite element analysis, and artificial neural network analysis of a complex nonlinear structure: a study on a long bone

Abstract

The bodies of humans and animals act like a mechanical system, in which bones are the major link for withstanding compressive loads. The extraordinarily hierarchical essence of the structure and mechanical properties of bones makes bone a difficult material to fully comprehend. It would be difficult to find another material that possesses such multivariable and exceptional mechanical properties. The horse has been identified as a useful model to investigate the effects of mechanical loading and induced deformations on tissues and structures in humans. Also, the equine distal forelimb is the source of most career-ending injuries to sport horses in all disciplines, and as such understanding how the materials in this part of the body respond to loads is key to understanding how and why they break down. The equine third metacarpal (MC3) is a large bone that shows a relatively restricted and simple range of normal movements in-vivo, has minimal muscle attachments (lacking any muscle or ligamentous attachments for a range around 6-8 cm along its midshaft), forms an essential part of the lower forelimb in withstanding loads, is vulnerable to failure during the majority of disastrous injuries occurring in racing horses worldwide, is well-adapted (well-engineered) for exercise at high speeds, and can withstand substantial compressive load as much as 22.5 kN in-vivo. To compare this huge force that MC3 is capable of withstanding in a normal daily basis, a large SUV car weighs 20.6 kN to 23.3 kN. Furthermore, this anatomical region provides a potential insight into how complex biological materials like bone and ligament respond to loading in all species and what parameters are vital to retain their integrity during normal use. Despite recent enormous advancement in techniques, gaining deep knowledge to find correlations between whole bone shape and material, mechanical, and overall physical responses as well as properties is a daunting task due to both complexity of the material itself and the convoluted shapes that this complex material forms. Moreover, many uncertainties and ambiguities exist concerning the use of computational techniques that unfortunately triggers a growing tendency for the misapplication of these computational tools. This thesis presents novel computational modelling frameworks and algorithms to build reliable 3D models of MC3 long bones and to replicate data recorded via ex-vivo experiments undertaken on MC3 bones. The major endeavor was to sufficiently establish precise and accurate computational models employing novel techniques (computer-aided design, finite element analysis, and machine learning algorithms using artificial neural networks) to predict the mechanical responses (for example, cyclic mechanical loading, displacement, and strains recorded experimentally) of a highly nonlinear well-engineered structure (MC3 long bone). This thesis reports the: (1) establishment of a reliable finite element analysis (FEA) and computer-aided design (CAD) model with reduced errors, (2) investigation of the significance of shape and geometry variations in the bone midshaft (where the strains are typically read) versus age and maturation, and (3) employment of novel machine learning algorithms with predominant use of feedforward back-propagation neural networks to tackle the enormous challenges in comprehending mechanical responses in long bones and equine forelimb mechanics. Otherwise conventional tools fail to provide a satisfactory outcome due to the availability of limited data and the extreme complexity of anticipating mechanical properties and discovering performance attributes in this highly interdisciplinary domain. Bones have remarkable properties and, as such, offer intriguing opportunities for investigation by researchers and engineers. The use of equine MC3 long bones in performing mechanical experiments and then in establishing the computational techniques in this thesis is motivational for engineers who are dealing with less complicated structures and less complex materials. The encouraging outcomes also exhibit the exceptional ability of artificial neural networks in capturing the mechanical characteristics of complex structures and will assist researchers and engineers as well as biologists in discerning mechanical features of complex engineering structures that exhibit exotic and peculiar nonlinear mechanical properties. These models and frameworks would potentially assist in improving clinical assessments, avoiding and/or minimizing injuries that inhibit performance of humans and animals, and encourage the early detection of associated injuries and failures.