Impact of Sampling Technique on the Performance of Surrogate Models Generated with Artificial Neural Network (ANN): A Case Study for a Natural Gas Stabilization Unit

• DOI: 10.3390/en12101906
• Publication Date: 2019-05-20T15:05:07Z
• Authors: Mohamed Ibrahim; Saad Al-Sobhi; Rajib Mukherjee; Ahmed AlNouss
• Keywords: Latin Hypercube Sampling; Surrogate model; Robustness

Data-driven models are essential tools for the development of surrogate that can be used design, operation, and optimization industrial processes. One approach developing is through use input–output data obtained from a process simulator. To enhance model robustness, proper sampling techniques required to cover entire domain variables uniformly. In present work, Monte Carlo with pseudo-random samples as well Latin hypercube quasi-Monte Hammersley Sequence Sampling (HSS) generated. The sampled simulator fitted neural networks generating model. An illustrative case study solved predict gas stabilization unit performance. From developed data, it concluded different methods, HSS have better performance than method designing This argument based on maximum absolute value, standard deviation, confidence interval relative average error techniques.