Data-driven optimization

Data from simulations or from process industries can be high-dimensional, sparse, uncertain, heterogeneous, multi-scale and represent discontinuous nonlinear functions. Novel methodology is required for data-driven optimization for applications in design, real-time optimization, scheduling, and process operations. This session seeks presentations on new mathematical optimization algorithms for data-driven optimization and/or applications to the process industries. Contributions may incorporate (i) model-free methods such as hardware-in-the-loop for process development, (ii) the development and use of surrogate models, (iii) methodologies for dealing with large-scale data sets, extracted from simulation or industrial historical data, and the information content in these data sets.