Inferential monitoring and optimization of crude separation units via hybrid models

Hybrid semi-parametric modeling in process systems engineering: Past, present and future

“…The vapor-liquid equilibrium is in this approach described by a RBFN. Similar approaches are also pro-posed by Mahalec and Sanchez (2012), Hinchliffe et al (2003) and Arahal et al (2008), where Hinchliffe et al (2003) divides the polymerization reactor into several stages whereas Arahal et al (2008) uses discrete volume and wall segments. The difficulty in the just named approaches is that for the training of the nonparametric model, sufficient data must be available, and that a nonparametric model trained with global data might not perform well locally.…”

“…The focus in this section is on modeling applications that deal with experimental data rather than data from virtual, simulated experiments, because simulation cases are typically applied to validate a pro-posed methodology (methodologies have been discussed above), while experimental studies are much more practically oriented. It can be seen in (Bhutani et al, 2006;Feil et al, 2004;Fiedler & Schuppert, 2008;Hinchliffe, Montague, Willis, & Burke, 2003;Mogk et al, 2002;Tian et al, 2001;Tsen et al, 1996;Vega, Lima, & Pinto, 2000), crystallization (Georgieva & de Azevedo, 2009;Georgieva, Meireles, & Feyo de Azevedo, 2003;Lauret, Boyer, & Gatina, 2000), metallurgic processes (Hu et al, 2011;Jia, Mao, Chang, & Zhao, 2011;Reuter et al, 1993;Sohlberg, 2005), distilla-tion columns (Chen et al, 2004;Mahalec & Sanchez, 2012;Safavi, Nooraii, & Romagnoli, 1999), drying processes (Cubillos & Acuna, 2007), thermal devices (Arahal, Cirre, & Berenguel, 2008), mechan-ical reactors (Nascimento et al, 1999) or milling (Aguiar & Filho, 2001; Kumar Akkisetty, Lee, Reklaitis, & Venkatasubramanian, 2010), for more details and references see Table2 -supplementary material. Since the number of applications is relatively large and most of which either use the standard serial approach consisting of material and/or energy balances in which the kinetics are represented by a nonparametric model or a parallel approach, only some approaches are discussed here, namely those that present solutions of more complex problems.…”

“…the white-box model predictions are considered as inputs to the nonparametric model or to establish a link between the process state and certain process characterizing parameters (Aguiar & Filho, 2001;Hwang et al, 2009;Mahalec & Sanchez, 2012;Nascimento, Giudici, & Scherbakoff, 1999;Quiza, Lopez-Armas, & Davim, 2012;Schenker & Agarwal, 2000;Zhang, Pan, Quan, Chen, & Shi, 2006). In general, hybrid structure C models did not find much application in chemical or biochemical engineer-ing, but in mechanical engineering (Quiza et al, 2012).…”

“…Hybrid semi-parametric models have been used to optimize the control policy either to maximize some quantity Eslamloueyan & Setoodeh, 2011;Henriques et al, 1999;Ignova et al, 2002;Kahrs & Marquardt, 2007;Mahalec & Sanchez, 2012;Preusting et al, 1996;Psichogios & Ungar, 1992;Schubert et al, 1994a;Teixeira et al, 2005Teixeira et al, , 2006Tholudur & Ramirez, 1996, 1999Zuo & Wu, 2000) or to meet specific quality specifications (Doyle et al, 2003;Hermanto et al, 2011;Safavi et al, 1999;Tian et al, 2001;Zhang et al, 2012) (see Table 10 -supplementary material for a complete list). Theoretically, the control policy can be optimized off-line or on-line.…”

“…Then, the optimization is constrained by a user defined risk (typically 25%) taking the minimal distance between the inputs obtained during the optimization and the closest cluster mean into account. Mahalec and Sanchez (2012) propose to constrain the optimization by two measures, one accounting for the distance of the current inputs to historical ones and a second ensuring that the residual and bias of the predictions in relation to the model plane do not exceed a certain threshold. Similarly, in Kahrs and Marquardt (2007) two complementary criteria to check the validity domain of hybrid semi-parametric models are proposed: (1) A convex-hull criteria to check whether each empirical model part only interpolates the data encountered during model identification; and (2) a confidence interval criterion with which the confidence intervals for the hybrid semi-parametric model are calculated.…”