Mathematical and Neural Approaches in Dependability Engineering: Study Case for a Technical System
Mathematical and Neural Approaches in Dependability Engineering: Study Case for a Technical System

Summary/Abstract: Predictive modeling with neural networks (NN) can be successfully applied in designing quantitative and qualitative parameters for the performance of technical systems. Neural network modeling enables design engineers in safety of operation to iteratively and interactively test the parameters and evaluate the corresponding changes in system performance before physical realization. For this reason, NN is an efficient tool in designing systems. NN can also be used during system testing as additional tools to determine the optimal operating values of the parameters and to determine their tolerance. The novelty of this paper is the combination into compact algorithm of mathematical and neural techniques in the realization of the correlations of the safety operation parameters and in the optimization of the dependability system in order to support the production-quality strategies. If mathematical modeling is the basis for the optimization operation, providing multivariable functions that connect parameters dependent on the independent parameters of a technical system, function that can give extreme points for variables, useful values in operational management, neural modeling creates the instrument for validating the mathematical model, and, through its structural flexibility and its ability to "learn" with new sets of models, can be used independently in the dependability decision process. The paper defines the problems in the field of operational safety, proposes relations of correlation between its parameters, defines the production-quality strategies, the mathematical model and the neural model and applies this complex technique to experimental data sets provided by a technical system. Validation of models is based on experimental patterns and conclusions have been drawn.

• Details
• Contents