Just-in-time learning based soft sensor with variable selection and weighting optimized by evolutionary optimization for quality prediction of nonlinear processes

A Moving Window Double Locally Weighted Extreme Learning Machine on an Improved Sparrow Searching Algorithm and Its Case Study on a Hematite Grinding Process

In this paper, a double locally weighted extreme learning machine model based on a moving window is developed to realize process modeling. To improve model performances, an improved sparrow-searching algorithm is proposed to optimize the parameters of the proposed model. The effectiveness of the proposed model and algorithm are verified by data from a hematite grinding process. The experimental results show that the proposed algorithm can significantly improve the global search ability and convergence speed in the parameter optimization of the proposed model. The proposed model can correctly estimate the grinding particle size which is expected to be applied to other complex industries.

Industrial Soft Sensor Optimized by Improved PSO: A Deep Representation-Learning Approach

Soft sensors based on deep learning approaches are growing in popularity due to their ability to extract high-level features from training, improving soft sensors' performance. In the training process of such a deep model, the set of hyperparameters is critical to archive generalization and reliability. However, choosing the training hyperparameters is a complex task. Usually, a random approach defines the set of hyperparameters, which may not be adequate regarding the high number of sets and the soft sensing purposes. This work proposes the RB-PSOSAE, a Representation-Based Particle Swarm Optimization with a modified evaluation function to optimize the hyperparameter set of a Stacked AutoEncoder-based soft sensor. The evaluation function considers the mean square error (MSE) of validation and the representation of the features extracted through mutual information (MI) analysis in the pre-training step. By doing this, the RB-PSOSAE computes hyperparameters capable of supporting the training process to generate models with improved generalization and relevant hidden features. As a result, the proposed method can generate more than 16.4% improvement in RMSE compared to another standard PSO-based method and, in some cases, more than 50% improvement compared to traditional methods applied to the same real-world nonlinear industrial process. Thus, the results demonstrate better prediction performance than traditional and state-of-the-art methods.

Deep Semi-Supervised Just-in-Time Learning Based Soft Sensor for Mooney Viscosity Estimation in Industrial Rubber Mixing Process

Soft sensor technology has become an effective tool to enable real-time estimations of key quality variables in industrial rubber-mixing processes, which facilitates efficient monitoring and a control of rubber manufacturing. However, it remains a challenging issue to develop high-performance soft sensors due to improper feature selection/extraction and insufficiency of labeled data. Thus, a deep semi-supervised just-in-time learning-based Gaussian process regression (DSSJITGPR) is developed for Mooney viscosity estimation. It integrates just-in-time learning, semi-supervised learning, and deep learning into a unified modeling framework. In the offline stage, the latent feature information behind the historical process data is extracted through a stacked autoencoder. Then, an evolutionary pseudo-labeling estimation approach is applied to extend the labeled modeling database, where high-confidence pseudo-labeled data are obtained by solving an explicit pseudo-labeling optimization problem. In the online stage, when the query sample arrives, a semi-supervised JITGPR model is built from the enlarged modeling database to achieve Mooney viscosity estimation. Compared with traditional Mooney-viscosity soft sensor methods, DSSJITGPR shows significant advantages in extracting latent features and handling label scarcity, thus delivering superior prediction performance. The effectiveness and superiority of DSSJITGPR has been verified through the Mooney viscosity prediction results from an industrial rubber-mixing process.

A novel two-step adaptive multioutput semisupervised soft sensor with applications in wastewater treatment

To make full use of unlabeled data for soft-sensor modelling and to address the coexistence of a large number of hard-to-measure variable issues, this study proposed a novel two-step adaptive heterogeneous co-training multioutput model. First, unlabeled data with the highest confidence were selected to optimize the model. Then, the proposed model co-trained Gaussian process regression (GPR) and least squares support vector machine (LSSVM) algorithms with two sets of independent labeled data. Second, at each step of the model update, the Kalman filter (KF) worked together with a moving window (MW) to strengthen the model to address process dynamics. Finally, the proposed model was demonstrated by a simulated wastewater treatment platform, BSM1, and a real sewage treatment plant. The root-mean-square error (RMSE) and root-mean sum of squares of the diagonal (RMSSD) were obviously reduced, and the correlation coefficient (R) and correlation coefficient (RR) reached 0.8 in both case studies. The results suggest that the proposed model can significantly improve prediction performance.

Industrial Semi-Supervised Dynamic Soft-Sensor Modeling Approach Based on Deep Relevant Representation Learning

Soft sensors based on deep learning have been growing in industrial process applications, inferring hard-to-measure but crucial quality-related variables. However, applications may present strong non-linearity, dynamicity, and a lack of labeled data. To deal with the above-cited problems, the extraction of relevant features is becoming a field of interest in soft-sensing. A novel deep representative learning soft-sensor modeling approach is proposed based on stacked autoencoder (SAE), mutual information (MI), and long-short term memory (LSTM). SAE is trained layer by layer with MI evaluation performed between extracted features and targeted output to evaluate the relevance of learned representation in each layer. This approach highlights relevant information and eliminates irrelevant information from the current layer. Thus, deep output-related representative features are retrieved. In the supervised fine-tuning stage, an LSTM is coupled to the tail of the SAE to address system inherent dynamic behavior. Also, a k-fold cross-validation ensemble strategy is applied to enhance the soft-sensor reliability. Two real-world industrial non-linear processes are employed to evaluate the proposed method performance. The obtained results show improved prediction performance in comparison to other traditional and state-of-art methods. Compared to the other methods, the proposed model can generate more than 38.6% and 39.4% improvement of RMSE for the two analyzed industrial cases.

Autoencoder-based nonlinear Bayesian locally weighted regression for soft sensor development

The framework of locally weighted learning (LWL) has established itself as a popular tool for developing nonlinear soft sensors in process industries. For LWL-based soft sensors, the key factor for achieving high performance is to construct accurate localized models. To this end, in this paper a nonlinear local model training algorithm called nonlinear Bayesian weighted regression (NBWR) is proposed. In the NBWR, the nonlinear features of process data are first extracted by the autoencoder; then, given a query sample a local dataset is selected on the feature space and a fully Bayesian regression model with differentiated sample weights is developed. The benefits of this approach, which include better consistency of correlation, stronger abilities to deal with process nonlinearities and uncertainties, overfitting and numerical issues, lead to superior performance. The NBWR is used for developing a soft sensor under the LWL framework, and a real-world industrial process is used to evaluate the performance of the NBWR-based soft sensor. The experimental results demonstrate that the proposed method outperforms several benchmarking soft sensing approaches.

Ensemble Just-In-Time Learning-Based Soft Sensor for Mooney Viscosity Prediction in an Industrial Rubber Mixing Process

The lack of online sensors for Mooney viscosity measurement has posed significant challenges for enabling efficient monitoring, control, and optimization of industrial rubber mixing process. To obtain real-time and accurate estimations of Mooney viscosity, a novel soft sensor method, referred to as multimodal perturbation- (MP-) based ensemble just-in-time learning Gaussian process regression (MP-EJITGPR), is proposed by exploiting ensemble JIT learning. This method employs perturbations on similarity measure and input variables for generating the diversity of JIT learners. Furthermore, a set of accurate and diverse JIT learners are built through an evolutionary multiobjective optimization by balancing the accuracy and diversity objectives explicitly. Moreover, all base JIT learners are combined adaptively using a finite mixture mechanism. The proposed method is applied to an industrial rubber mixing process for Mooney viscosity prediction, and the experimental results demonstrate its effectiveness and superiority over traditional soft sensor methods.

Evaluating the Performance of Newly Integrated Model in Nonlinear Chemical Process Against Missing Measurements

Application of data-driven soft sensors in manufacturing fields, for instance, chemical, pharmaceutical, and bioprocess have rapidly grown. The issue of missing measurements is common in chemical processing industries that involve data-driven soft sensors. Locally weighted Kernel partial least squares (LW-KPLS) algorithm has recently been proposed to develop adaptive soft sensors for nonlinear processes. This algorithm generally works well for complete datasets; however, it is unable to cope well with any datasets comprising missing measurements. Despite the above issue, limited studies can be found in assessing the effects of incomplete data and their treatment method on the predictive performances of LW-KPLS. To address these research gaps, therefore, a trimmed scores regression (TSR) based missing data imputation method was integrated to LW-KPLS to formulate trimmed scores regression assisted locally weighted Kernel partial least squares (TSR-LW-KPLS) model. In this study, this proposed TSR-LW-KPLS was employed to deal with missing measurements in nonlinear chemical process data. The performances of TSR-LW-KPLS were evaluated using three case studies having different percentages of missing measurements varying from 5 % to 40 %. The obtained results were then compared to the results from singular value decomposition assisted locally weighted Kernel partial least squares (SVD-LW-KPLS) model. SVD-LW-KPLS was also proposed by incorporating a singular value decomposition (SVD) based missing data treatment method into LW-KPLS. From the comparative studies, it is evident that the predictive accuracies of TSR-LW-KPLS are superior compared to the ones from SVD-LW-KPLS.