Soft sensors have been increasingly applied for quality prediction in complex industrial processes, which often have different scales of topology and highly coupled spatiotemporal features. However, the existing soft sensing models usually face difficulties in extracting the multiscale local spatiotemporal features in multicoupled complex process data and harnessing them to their full potential to improve the prediction performance. Therefore, a multiscale attention-based CNN (MSACNN) is proposed in this article to alleviate such problems. In MSACNN, convolutional kernels of different sizes are first designed in parallel in the convolutional layers, which can generate feature maps containing local spatiotemporal features at different scales. Meanwhile, a channel-wise attention mechanism is designed on the feature maps in parallel to get their attention weights, representing the significance of the local spatiotemporal feature at different scales. The superiority of the proposed MSACNN over the other state-of-the-art methods is validated through the performance evaluation in two real industrial processes.

In complex process industries, multivariate time sequences are omnipresent, whose nonlinearities and dynamics present two major challenges for soft sensing of important quality variables. Consequently, due to the potent representational capabilities, nonlinear dynamic models like gated recurrent unit (GRU) and long short-term memory (LSTM) networks have been used for data sequence modeling. Though it is a common occurrence in many industrial plants, data series with heterogeneous sample intervals and missing values cannot be directly handled by these dynamic algorithms. To this end, attention-based interval-aided networks (AIA-Net) are proposed in this article to adaptively model the temporal information for heterogeneous sampling sequences with missing values in the processes industry. It includes two main mechanisms, which are named attention-based time-aware dynamic imputation and interval-aided time-aware network, respectively. The reduction rate is introduced by the attention-based time-aware dynamic imputation to apply the effects of time intervals and is used in the imputation of missing data. The interval-aided time-aware network includes time intervals in the model structure and uses a sampling interval gate to correct the temporal correlations in time series. The proposed AIA-Net is successfully applied to a real hydrocracking process to predict the C5 and C6 content in the light naphtha.

The growth of data collection in industrial processes has led to a renewed emphasis on the development of data-driven soft sensors. A key step in building an accurate, reliable soft sensor is feature representation. Deep networks have shown great ability to learn hierarchical data features using unsupervised pretraining and supervised fine-tuning. For typical deep networks like stacked auto-encoder (SAE), the pretraining stage is unsupervised, in which some important information related to quality variables may be discarded. In this article, a new quality-driven regularization (QR) is proposed for deep networks to learn quality-related features from industrial process data. Specifically, a QR-based SAE (QR-SAE) is developed, which changes the loss function to control the weights of the different input variables. By choosing an appropriate inductive bias for the weight matrix, the model provides quality-relevant information for predictive modeling. Finally, the proposed QR-SAE is used to predict the quality of a real industrial hydrocracking process. Comparative experiments show that QR-SAE can extract quality-related features and achieve accurate prediction performance.

Soft sensors have been extensively developed to estimate the difficult-to-measure quality variables for real-time process monitoring and control. Process nonlinearities and dynamics are two main challenges for accurate soft sensor modeling. To cope with these problems, two temporal convolutional network (TCN)-based soft sensor models are developed in this paper. With a hierarchy of temporal convolution kernels and large receptive fields, TCN is able to describe the nonlinearities and long dynamic dependence of process variables. Thus, a TCN model is first designed for nonlinear dynamic soft sensing. However, the original TCN model neglects the auto-correlations of the quality variable, as well as the cross-correlations between the quality variable and process variables. Hence, an autoregressive TCN (AR-TCN) is further constructed by integrating the lagged quality data in the past neighboring samples for quality prediction of the current sample. Thus, AR-TCN is able to capture the auto-correlations and cross-correlations between the quality variables and process variables, which are more beneficial for quality prediction. The effectiveness of two proposed models is validated on an industrial debutanizer column and an industrial hydrocracking process.

In process industries, accurate prediction of critical quality variables is particularly important for process control and optimization. Usually, soft sensors have been developed to estimate the quality variables via process variables. However, there are often process variables that are far apart in topology but have high correlations, presenting challenges for such feature learning in soft sensor. To solve these difficulties, a variable correlation analysis based convolutional neural network (VCA-CNN) is proposed in this paper for far topological feature extraction, which generates three-channel data samples containing natural topology features, linear topology features, and nonlinear topology features through Pearson and Spearman correlation to adequately focus on far-topology features. In the feature extraction stage, the VCA-CNN employs a special row convolution kernel to extract far-topological features. Then, normal convolutional and pooling operations are alternately processed on the output feature maps. This approach overcomes the limitation that traditional convolutional operations mainly extract local correlations of process variables. The effectiveness and superiority of the VCA-CNN are validated in the hydrocracking process and debutanizer column.

Missing values are a common occurrence in industrial datasets, resulting from multiple sampling rates, sensor malfunctions, and transmission errors, whose presence can significantly affect the accuracy of data-driven models. An effective method to solve this problem is to impute the missing data in advance. This paper proposes a new position-encoding denoising auto-encoder (PE-DAE), which is motivated by the advantages of DAE in data reconstruction. To make use of the known information, the autocorrelation of the time-series data and the information on missing positions are considered. Moreover, a self-paced learning (SPL) training strategy is proposed to improve the imputation performance under different levels of missing data. The SPL training framework can first learn the knowledge structure of data with low missing rates, and then gradually increase the difficulty, transitioning to learning more complex knowledge from data with higher missing rates. Finally, the proposed method is used for missing data imputation in two real industrial processes. Comparative experiments show that the PE-DAE+SPL achieves the smallest error at all missing rates.