In modern industrial processes, soft sensors have played an important role for effective process control, optimization, and monitoring. Feature representation is one of the core factors to construct accurate soft sensors. Recently, deep learning techniques have been developed for high-level abstract feature extraction in pattern recognition areas, which also have great potential for soft sensing applications. Hence, deep stacked autoencoder (SAE) is introduced for soft sensor in this paper. As for output prediction purpose, traditional deep learning algorithms cannot extract high-level output-related features. Thus, a novel variable-wise weighted stacked autoencoder (VW-SAE) is proposed for hierarchical output-related feature representation layer by layer. By correlation analysis with the output variable, important variables are identified from other ones in the input layer of each autoencoder. The variables are assigned with different weights accordingly. Then, variable-wise weighted autoencoders are designed and stacked to form deep networks. An industrial application shows that the proposed VW-SAE can give better prediction performance than the traditional multilayer neural networks and SAE.

Industrial process data are naturally complex time series with high nonlinearities and dynamics. To model nonlinear dynamic processes, a long short-term memory (LSTM) network is very suitable for soft sensor model development. However, the original LSTM does not consider variable and sample relevance for quality prediction. In order to overcome this problem, a spatiotemporal attention-based LSTM network is proposed for soft sensor modeling, which can, not only identify important input variables that are related to the quality variable at each time step, but also adaptively discover quality-related hidden states across all time steps. By taking the spatiotemporal quality-relevant interactions into consideration, the prediction performance can be improved for the soft sensor model. The effectiveness and flexibility of the proposed model is demonstrated on an industrial hydrocracking process to predict the initial boiling points of heavy naphtha and aviation kerosene.

Deep learning is a recently developed feature representation technique for data with complicated structures, which has great potential for soft sensing of industrial processes. However, most deep networks mainly focus on hierarchical feature learning for the raw observed input data. For soft sensor applications, it is important to reduce irrelevant information and extract quality-relevant features from the raw input data for quality prediction. To deal with this problem, a novel deep learning network is proposed for quality-relevant feature representation in this paper, which is based on stacked quality-driven autoencoder (SQAE). First, quality-driven autoencoder (QAE) is designed by exploiting the quality data to guide feature extraction with the constraint that the potential features should largely reconstruct the input layer data and the quality data at the output layer. In this way, quality-relevant features can be captured by QAE. Then, by stacking multiple QAEs to construct the deep SQAE network, SQAE can gradually reduce irrelevant features and learn hierarchical quality-relevant features. Finally, the high-level quality-relevant features can be directly applied for soft sensing of the quality variables. The effectiveness and flexibility of the proposed deep learning model are validated on an industrial debutanizer column process.

Industrial process plants are instrumented with a large number of redundant sensors and the measured variables are often contaminated by random noises. Thus, it is significant to discover the general trends of data by latent variable models in the probabilistic framework before soft sensor modeling. However, traditional probabilistic latent variable models such as probabilistic principal component analysis are mostly static linear approaches. The process dynamics and nonlinearities have not been well considered. In this paper, a novel weighted linear dynamic system (WLDS) is proposed for nonlinear dynamic feature extraction. In WLDS, two kinds of weights are proposed for local linearization of the nonlinear state evolution and state emission relationships. In this way, a weighted log-likelihood function is designed and expectation-maximization algorithm is then used for parameter estimation. The feasibility and effectiveness of the proposed method is demonstrated with a numerical example and an industrial process application.