Global sensitivity analysis for identifying critical process design decisions

• A methodology to avoid unwanted responses in the designed process is presented.
• Sobol’ method is used to eliminate non-influential input variables.
• Monte Carlo Filtering is used for determination and regionalization of critical variables.
• Mineral concentration circuit and RO desalination plant designs are used as examples.
• The methodology could be helpful to consider design under uncertainty.

Process design is performed using deterministic values of input variables. However, these input variables may have uncertainties that can lead the designed process to unwanted conditions. This paper presents a methodology for determining critical variables to avoid unwanted responses in the designed process. The methodology proposed here consists of three stages: (1) deterministic process design, (2) elimination of non-influential input variables via the global sensitivity analysis method of Sobol’, and (3) determination and regionalization of critical variables with Monte Carlo Filtering. The proposed methodology was applied to the design of a mineral concentration circuit and to the design of a desalination plant. The results show that the methodology could be helpful in analyzing whether a deterministic design works adequately under uncertainties, identifying critical variables in large models, and regionalizing operating conditions and design variables.

Figure optionsDownload high-quality image (76 K)Download as PowerPoint slide