Quality Assurance of Mechanical Systems and Structures

The development of reliable products is one of the most important prerequisites for industrial nations like Austria to maintain a competitive edge in the international competition. Higher complexity, a more extensive functionality, increasing product reliability, decreasing costs for development, shorter periods for development, etc. became the key issues in product development. Requirements for guaranteeing certain quality levels e.g. in terms of a set target reliability etc. - and at the same time minimizing the production costs have to be met. The required high reliability of mechanical systems nowadays can not be guaranteed by still relying solely on the classical way of design methods which are based on long term experience only. Nowadays analytical and numerical reliability methods have to be used to fulfill the increased demands. For this qualitative, and even more important, quantitative methods of reliability analysis have to be applied as part of reliability management. However, the current state of the engineering practice,-aside from some special types of investigations- still does not make use of these modern methods of reliability-based analysis with which the above stated requirements can be met. Currently, structural analysis in all types of engineering, e.g. a civil, mechanical automotive, aerospace, chemical engineering etc. is still carried out applying traditional deterministic analyses. All known and unknown uncertainties in loading, geometric material properties, etc. are taken into account then in the design process by applying mainly empirically based safety factors. In other words the uncertainties are not treated rationally where they occur, but are lumped into global or partial safety factors. It is only for the rational treatment of uncertainties in the structural analysis which allows an economical design and the performance of quality assurance of mechanical systems. This fact has already been recognized more than half a century ago and since then the scientific developments in the areas of stochastic structural analysis and reliability have been quite dramatic. Yet, the developments remained still short of its large scale practical application. Too many steps are still missing until software for reliability based analysis- similarly successful as the finite element software- could be developed in such a way that it disseminates into day-to-day structural analysis and hence becomes an indispensable part of it. The necessary steps for bringing the current scientific developments to such a state that they can be picked up by the industry for large scale use have to be performed and are proposed in this research project.

One of the main goals of structural engineering is to develop safe and efficient designs by employing available methods and technology. Accurate prediction of the performances of structures via mathematical models plays a crucial role in this regard, mainly because they reduce the costs of designing, producing, operating and maintaining processes dramatically. Particularly when new systems have to be developed and designed, the traditional repeated testing procedures are much too time consuming and expensive. Hence, the development products nowadays are mostly carried out by sophisticated mechanical modeling and by almost exclusively using computers. However, this also requires methodologies so that the inherent uncertainties, which in fact are observed in traditional experimental design procedures, can be assessed and processed. Only this way a quality assurance of the products is feasible. Methods of stochastic analysis, in this regard, are capable of accomplishing this task. While stochastic methods have been successfully developing in the research field, a large portion of the potential of these methods has not been released yet. This is mainly due to the missing link between the methodologies developed in the basic research arena and the stage at which industry is eager to adopt the methods. Therefore it has been the objective of this project to fill in this gap between basic research and the applications. A general purpose and user-friendly software, namely COSSAN (COmputational Stochastic Structural ANalysis), has been developed for this purpose. More specifically, the developed software offers advanced algorithms to perform various analysis types, such as Sensitivity/Robustness Analysis, Uncertainty Quantification, Reliability Analysis, Reliability based Optimization and Life Cycle Management. In other words, a large spectrum of capabilities has been integrated into a single package, which is designed to be easily used by practitioners without extensive training. The software is also equipped with advanced computational tools, such as parallel computation algorithms, in order to handle analysis of large and complex systems. In summary, the project resulted in the translation of the developments within the research field into the industrial applications. The long-term benefits to the engineering community are foreseen as unleashing the great potential of stochastic mechanics by offering the necessary tools to the industry. Finally, it is also envisioned that as the remedies of stochastic analysis are recognized by the authorities, these methods will slowly but surely become a routine part of the design cycle within engineering practice.