RT info:eu-repo/semantics/article
T1 Knowledge based recursive non-linear partial least squares (RNPLS)
A1 Merino, Alejandro
A1 García Álvarez, Diego
A1 Sáinz Palmero, Gregorio Ismael
A1 Acebes Arconada, Luis Felipe
A1 Fuente Aparicio, María Jesús de la
K1 Soft sensor
K1 Partial Least Squares
K1 Non-linear mapping
K1 Recursive estimation
K1 RNPLS
AB Soft sensors driven by data are very common in industrial plants to perform indirect measurementsof difficult to measure critical variables by using other variables that are relatively easier to obtain.The use of soft sensors implies some challenges, such as the colinearity of the predictor variables,the time-varying and possible non-linear nature of the industrial process. To deal with the firstchallenge, the partial least square (PLS) regression has been employed in many applications to modelthe linear relations between process variables, with noisy and highly correlated data. However, thePLS model needs to deal with the other two issues: the non-linear and time-varying characteristicsof the processes. In this work, a new knowledge-based methodology for a recursive non-linear PLSalgorithm (RNPLS) is systematized to deal with these issues. Here, the non-linear PLS algorithm is setup by carrying out the PLS regression over the augmented input matrix, which includes knowledgebased non-linear transformations of some of the variables. This transformation depends on the system’snature, and takes into account the available knowledge about the process, which is provided by expertknowledge or emulated using software tools. Then, the recursive exponential weighted PLS is used tomodify and adapt the model according to the process changes. This RNPLS algorithm has been testedusing two case studies according to the available knowledge, a real industrial evaporation station ofthe sugar industry, where the expert knowledge about the process permits the formulation of therelationships, and a simulated wastewater treatment plant, where the necessary knowledge about theprocess is obtained by a software tool. The results show that the methodology involving knowledgeregarding the process is able to adjust the process changes, providing highly accurate predictions.
PB Elsevier
SN 0019-0578
YR 2020
FD 2020
LK http://uvadoc.uva.es/handle/10324/45595
UL http://uvadoc.uva.es/handle/10324/45595
LA eng
NO ISA Transactons, Mayo 2020, vol. 100, p.481-494
NO Producción Científica
DS UVaDOC
RD 18-may-2024