Soft sensor based on Gaussian process regression and its application in erythromycin fermentation process

Abstract

Article Highlights• A systematic soft sensor modeling method based on GPR and PCA is proposed • The variance of the predicted output was designed on the output uncertainty of the GPR model • Practical applications show the superiority of the proposed soft sensor method Abstract Erythromycin fermentation is a typical microbial fermentation process. Soft sensors can be used to estimate the biomass of Erythromycin fermentation process due to their relative low cost, simple development, and ability to predict difficult-to-measure variables. However, traditional soft sensors, e.g. artificial neural network (ANN) soft sensors, support vector machine (SVM) soft sensors, etc., cannot represent the uncertainty (measurement precision) of outputs, which results in difficulties in practice. Gaussian process regression (GPR) provides a novel framework to solve regression problems. The output uncertainty of a GPR model follows Gaussian distribution, expressed in terms of mean and variance. The mean represents the predicted output. The variance can be viewed as the measure of confidence in the predicted output that distinguishes the GPR from NN and SVM soft sensor models. We propose a systematic approach based on GPR and principal component analysis (PCA) to establish a soft sensor to estimate biomass of erythromycin fermentation process. Simulations on industrial data from an erythromycin fermentation process show the proposed GPR soft sensor has high performance of modeling the uncertainty of estimates.

Modern Soft-Sensing Modeling Methods for Fermentation Processes

“… Dual learning-based online ensemble regression approach for adaptive soft-sensing modeling of nonlinear time-varying processes [117].  A soft-sensing model based on GPR for the erythromycin fermentation process is presented in [118].  A soft-sensing modeling method based on multi-model strategy by using GPR and PCA is presented in [119].…”

Solid state treatment with Lactobacillus paracasei subsp. paracasei BGHN14 and Lactobacillus rhamnosus BGT10 improves nutrient bioavailability in granular fish feed

“…Normality of data in control and treatment groups was assessed by Shapiro-Wilk test. Since process modeling could not be performed by least square method, due to significant lack of fit, modeling was performed using Gaussian process regression (GPR) [33, 34], which does not have assumptions concerning the regression function distribution. Bayesian information criterion (BIC) and Pearson correlation coefficient between predicted and actual amounts for each analyzed parameter were calculated in order to evaluate model validity.…”

From Spatial-Temporal Multiscale Modeling to Application: Bridging the Valley of Death in Industrial Biotechnology

“…The quantification of the uncertainty makes GP a powerful tool within biological process control. For example, Mei et al established a soft-sensor model for erythromycin fermentation at an industrial scale with principal component analysis (PCA) which focused on selecting important features to simplify the model structure [ 95 ]. This method has excellent prediction performance for biomass concentration in the exponential growth period of erythromycin fermentation [ 95 ].…”

“…For example, Mei et al established a soft-sensor model for erythromycin fermentation at an industrial scale with principal component analysis (PCA) which focused on selecting important features to simplify the model structure [ 95 ]. This method has excellent prediction performance for biomass concentration in the exponential growth period of erythromycin fermentation [ 95 ]. Zhang et al optimized the process in phycocyanin production by cyanobacteria in a semi-batch bioreactor through nonlinear model predictive control (NMPC) and successfully constructed the optimal nitrate feed strategy in the actual plant production process [ 96 ].…”