New modelling techniques for dependability. Case study for a mechanical process

Abstract

Dependability - a measure of a system's availability, reliability, its maintainability and safety systems, is an integral part of the integrity of engineering design. From a functional perspective, dependability assumes satisfying some of the required functions and implies the correct definition of them. Thus, in the case of complex equipment, multi-variable functions can be generated whose variables are dependent on different states and operating modes. In recent years, automated continuous design techniques have been developed through intelligent computational methodologies to determine the dependability of systems. These techniques also introduce Artificial Intelligence (AI) in integrity engineering design and offer an overview and modelling in designing for reliability, availability, maintainability and system security. This paper combines quantitative research, mathematical and algorithmic modelling with evolutionary techniques. The combination of the four research techniques is done in a coherent application validated on a mechanical device and applicable to a wide range of technical systems. More specifically, the work proposes a comparative study of mathematical - linear and nonlinear models for correlation and optimization of maintainability and reliability values in relation to production and quality policies. These models will be complemented with specific AI techniques: neural and genetic algorithms that validate and optimize mathematical models. It identifies the metrics for the evaluation of the dependent variables (reliability, maintainability) and an independent variables (production and quality), these variables are observed over a given time horizon, their mathematical, neural and genetic modelling is performed to achieve correlations, optimization functions and model validation. The study proves to be a practical and applicative tool in the analysis of system integrity in general and operational safety, especially with a large generalization capacity on systems in many technical fields.