Soft sensors are essential for advanced monitoring and control to prevent undesirable operations and improve product quality. However, nonlinear, autocorrelated, and cross-correlated behaviors in industrial data demand concurrent modeling of the dynamics and nonlinearities. Deep learning-based soft sensors, such as recurrent neural network (RNN) and long short-term memory (LSTM) networks, often incorporate complex structures and numerous parameters, which can lead to an overly complex model. In practical applications where training data samples are limited, a lightweight neural network with strong generalization capability is preferred. With a simple structure of feed-forward layers of 1-D convolutional neural networks (CNNs) (1-D-CNN) for time-series data modeling, this article proposes a novel lightweight dynamic CNN (LDCNN) for soft sensors. Positional embedding (PE) and simplified temporal attention mechanisms are integrated for improved dynamic modeling, while dilated convolutions and layer normalization (LN) are incorporated to significantly reduce the depth and width of the network and avoid over-parametrization. Experimental results on a real industrial case indicate that a lightweight model outperforms the traditional methods with limited training samples.

The successive switching nature of multimode processes, coupled with data scarcity, challenges traditional quality prediction models. Specifically, the difficulty of simultaneously collecting abundant labeled datasets from all modes forces the model to update its parameters as modes switch. This leads to the forgetting of historical mode knowledge and hinders the aggregation of knowledge, thereby degrading generalization across modes. To this end, we propose a novel continual semisupervised graph echo state network ($\text{CS}^{2}$GESN). First, a semisupervised graph echo state network ($\text{S}^{2}$GESN) is designed based on the graph smoothing assumption to extract dynamic information from unlabeled samples within each mode. The$\text{S}^{2}$GESN model then evolves into a continual model,$\text{CS}^{2}$GESN, employing an elastic weight consolidation strategy for parameter importance estimation derived from pseudoinverse parameter optimization, facilitating the accumulation of historically learned knowledge. This manner alleviates performance deterioration from data scarcity and information forgetting, and enables more flexible modeling of successive arriving operating modes. The superiority and feasibility of the proposed method are demonstrated through its application to the Tennessee Eastman process and the three-phase flow facility process.

Conventional data-driven dynamic process monitoring methods usually rely on data collected at a single sampling rate. The effectiveness of these approaches typically diminishes when analyzing data from multiple sampling rates. To address this gap, this article introduces a new total multirate linear Gaussian state space model. This model is designed for modeling and monitoring in dynamic processes that involve data from various sampling rates. It works by establishing global dynamic latent variables that span across process variables and extracting local static latent variables for each sampling rate. For effective fault detection at different sampling rates, the model incorporates three kinds of statistics. The effectiveness of the proposed method in process monitoring is validated using the multiphase flow facility benchmark and a real papermaking wastewater treatment process.

Supervised cross-modal hashing has received wide attention in recent years. However, existing methods primarily rely on sample-wise semantic relationships to evaluate the semantic similarity between samples, overlooking the impact of label distribution on enhancing retrieval performance. Moreover, the limited representation capability of traditional dense hash codes hinders the preservation of semantic relationship. To overcome these challenges, we propose a new method, Joint Semantic Preserving Sparse Hashing (JSPSH). Specifically, we introduce a new concept of cluster-wise semantic relationship, which leverages label distribution to indicate which samples are more suitable for clustering. Then, we jointly utilize sample-wise and cluster-wise semantic relationships to supervise the learning of hash codes. In this way, JSPSH preserves both kinds of semantic relationships to ensure that more samples with similar semantics are clustered together, thereby achieving better retrieval results. Furthermore, we utilize high-dimensional sparse hash codes that offer stronger representation capability to preserve such more complex semantics. Finally, an interaction term is introduced in hash functions learning stage to further narrow the gap between modalities. Experimental results on three large-scale datasets demonstrate the effectiveness of JSPSH in achieving superior retrieval performance. Codes are available at https://github.com/hutt94/JSPSH.

The nonlinear and dynamic nature of complex industrial processes presents a significant challenge for monitoring incipient faults. To this end, this article proposes a novel deep dynamic latent variable model called dynamic inner canonical variate network (DiCVNet). The developed DiCVNet, which is in an end-to-end learning framework, consists of a dual convolutional autoencoder (DuCAE) and an autoregressive (AR) module. First, the DuCAE architecture with an AR module is designed to extract two correlated and self-orthogonal nonlinear dynamic canonical variables (CVs) from past and future datasets for tiny variation modeling. The AR module is embedded to extract the CVs with consistent dynamics for enhanced dynamic modeling of DuCAE. Then, a new incipient fault monitoring scheme for nonlinear dynamic processes is established. Finally, the performance of the proposed method is verified by two industrial cases, that are, a continuous stirred tank reactor and a multiphase flow process.

Traditional data-driven quality prediction methods are mainly built from static models using clean data with a slow sampling rate, leaving the process dynamics unused. To make full use of dynamic process data collected at a fast sampling rate, this article proposes a novel deep learning-based robust dual-rate dynamic data modeling method for quality prediction of dynamic nonlinear processes. A new dynamic data denoising generative adversarial imputation network is first proposed for the missing value imputation among the dynamic process data. Then, a new hint convolutional neural network (HCNN) is established for dual-rate data based quality prediction. The proposed HCNN incorporates the information hint mechanism of channel expansion into the convolutional neural network to extract the dynamic features with definitive time and variable information. Finally, the proposed method is verified using the Dow distillation process dataset and Beijing multisite air quality dataset.

Process shift of multimode process involving data distribution and dynamic relation makes traditional transfer learning methods be intractable and even result in negative transfer. To tackle this issue, this article proposes a novel self-tuning transfer dynamic modeling method for quality prediction of multimode processes. First, in order to capture domain-invariant spatiotemporal (DIST) features, a transfer dynamic convolution autoencoder (TDCAE) with a feature decomposition structure is established. Meanwhile, a first-order vector autoregressive constraint is embedded to extract consistent inner dynamics for DIST features. Then, a shared regression network is established to extract the relations with quality variables. Furthermore, by making full use of private spatiotemporal information from target labeled samples in response to the process shift, the self-tuning TDCAE (STDCAE) aided by a fine-tuning strategy is established for online compensation. Finally, the efficacy of the proposed TDCAE and STDCAE is demonstrated by a comprehensive study of a three-phase flow facility process.

Quality prediction is beneficial to intelligent inspection, advanced process control, operation optimization, and product quality improvements of complex industrial processes. Most of the existing work obeys the assumption that training samples and testing samples follow similar data distributions. The assumption is, however, not true for practical multimode processes with dynamics. In practice, traditional approaches mostly establish a prediction model using the samples from the principal operating mode (POM) with abundant samples. The model is inapplicable to other modes with a few samples. In view of this, this article will propose a novel dynamic latent variable (DLV)-based transfer learning approach, called transfer DLV regression (TDLVR), for quality prediction of multimode processes with dynamics. The proposed TDLVR can not only derive the dynamics between process variables and quality variables in the POM but also extract the co-dynamic variations among process variables between the POM and the new mode. This can effectively overcome data marginal distribution discrepancy and enrich the information of the new mode. To make full use of the available labeled samples from the new mode, an error compensation mechanism is incorporated into the established TDLVR, termed compensated TDLVR (CTDLVR), to adapt to the conditional distribution discrepancy. Empirical studies show the efficacy of the proposed TDLVR and CTDLVR methods in several case studies, including numerical simulation examples and two real-industrial process examples.

With excellent feature representation capabilities, deep autoencoder networks have attracted attention in process monitoring. However, it cannot take into account the quality indicators to identify whether the faults are quality-relevant. To address this issue, an orthogonal feature separation autoencoder (OFSAE) method is developed for quality-relevant fault monitoring. The proposed OFSAE mainly consists of the quality-relevant encoder network, quality-irrelevant encoder network, decoder network, and regression network. Through parallel learning and orthogonal projection for process variables, quality-relevant and quality-irrelevant variations can be isolated while maintaining good prediction performance. Finally, in comparison with conventional monitoring methods, the superiority of OFSAE is validated by the Tennessee Eastman process.

Intelligent manufacturing has promoted the development of Industry 4.0 and enabled the manufacturing industry to gradually move into the stage of intelligence with the rapid development of the Internet of Things and the Industrial Internet. An intelligent manufacturing system is a manufacturing system that can automatically adapt to changing environments and varying process requirements with minimal supervision and assistance from operators. Therefore, intelligent manufacturing has become a recognized core high technology to enhance the overall competitiveness of the manufacturing industry. The goal of intelligent manufacturing is to make production resources form a circular network with the characteristics of autonomy, adjustability, and configurability, to develop production processes flexibly, and to realize the efficiency of individual customization. For example, by analyzing the factory floor data, equipment monitored data, and the enterprise manufacturing database, it could help to store, explore, and make complex decisions for the manufacturing system. To achieve this goal, modern information technologies, such as artificial intelligence, big data, cloud computing, and mobile Internet, modeling, control, and optimization need to be integrated and collaborated with the physical resources of the manufacturing process, which triggers new theory, solution algorithms, and application scenarios.

In transfer learning model, the source domain samples and target domain samples usually share the same class labels but have different distributions. In general, the existing transfer learning algorithms ignore the interclass differences and intraclass similarities across domains. To address these problems, this article proposes a transfer learning algorithm based on discriminative Fisher embedding and adaptive maximum mean discrepancy (AMMD) constraints, called discriminative Fisher embedding dictionary transfer learning (DFEDTL). First, combining the label information of source domain and part of target domain, we construct the discriminative Fisher embedding model to preserve the interclass differences and intraclass similarities of training samples in transfer learning. Second, an AMMD model is constructed using atoms and profiles, which can adaptively minimize the distribution differences between source domain and target domain. The proposed method has three advantages: 1) using the Fisher criterion, we construct the discriminative Fisher embedding model between source domain samples and target domain samples, which encourages the samples from the same class to have similar coding coefficients; 2) instead of using the training samples to design the maximum mean discrepancy (MMD), we construct the AMMD model based on the relationship between the dictionary atoms and profiles; thus, the source domain samples can be adaptive to the target domain samples; and 3) the dictionary learning is based on the combination of source and target samples which can avoid the classification error caused by the difference among samples and reduce the tedious and expensive data annotation. A large number of experiments on five public image classification datasets show that the proposed method obtains better classification performance than some state-of-the-art dictionary and transfer learning methods. The code has been available at https://github.com/shilinrui/DFEDTL.

Complex industrial processes are commonly characterized with multiple operation modes. The existing manifold learning-based process monitoring methods describe each mode individually without capturing the connections among different modes, which may deteriorate the monitoring capability. This article proposes a novel dimensionality reduction model referred as to multimanifold joint projections to monitor the multimode processes, where the intramode and the intermode adjacency matrices are constructed to reflect the underlying features within each mode and among different modes, respectively. The neighboring and nonneighboring structures of data within each mode are captured by the distance and angle information of pairwise points to reveal the intrinsic structure of the original data, thus offering a more faithful representation of multimodal data and further enhancing monitoring performance. During online monitoring, a point to manifold distance criterion is proposed to determine the running-on mode of new samples. Two case studies demonstrated the superior performance of the proposed approach in multimode process monitoring.

This paper proposes a reference vector based multidirectional prediction strategy to address dynamic multiobjective optimization problems (DMOPs), whose PSs rotate with time. In this strategy, several reference vectors partition the population into multiple clusters and the same prediction trajectory is adapted in a cluster. In addition, a new offspring creation strategy in genetic operator named reference vector based offspring creation is proposed to accelerate the convergence and maintain the diversity. Experiments on seven benchmark problems are carried out to examine the performance of the proposed algorithm and the statistical results shows the proposed algorithm can address DMOPs with rotating PSs.

In this paper, a dynamic modeling method for unevenly sampled data is proposed for the monitoring of bi-layer (i.e., a process layer and a quality layer) dynamic processes. First, a novel uneven data dynamic canonical correlation analysis method with an integrated dynamic time window is proposed for interlayer latent structure modeling, which captures the dynamic relations between regularly sampled process data and quality data with slow and irregular sampling. The new model is a step toward big data modeling to deal with data irregularity and diversity. Second, after extracting covariations using an interlayer model, intralayer variations are extracted using subsequent principal component analysis on the residual subspaces of the original process data and quality data, respectively. Third, a concurrent monitoring method for unevenly sampled bi-layer data is proposed. Finally, the proposed method is demonstrated using an illustrative simulation example and applied successfully to a real blast furnace iron-making process.

Projection to latent structures (PLS) and concurrent PLS are approaches for solving quality-relevant process monitoring. In this paper, a new approach called concurrent kernel PLS (CKPLS) is presented to detect faults comprehensively for nonlinear processes. The new model divides the nonlinear process and quality spaces into five subspaces: the co-varying, process-principal, process-residual, quality-principal, and quality-residual subspaces. The co-varying subspace reflects nonlinear relationship between quality variables and original process variables. The process-principal and process-residual subspaces reflect the principal variations and residuals, respectively, in the nonlinear process space. Further, the quality-principal and quality-residual subspaces reflect the principal variations and residuals, respectively, in the quality space. The proposed approach is demonstrated by a numerical simulation and an application of the Tennessee Eastman process.

Process monitoring and fault diagnosis of the continuous annealing process lines (CAPLs) have been a primary concern in industry. Stable operation of the line is essential to final product quality and continuous processing of the upstream and downstream materials. In this paper, a multilevel principal component analysis (MLPCA)-based fault diagnosis method is proposed to provide meaningful monitoring of the underlying process and help diagnose faults. First, multiblock consensus principal component analysis (CPCA) is extended to MLPCA to model the large scale continuous annealing process. Secondly, a decentralized fault diagnosis approach is designed based on the proposed MLPCA algorithm. Finally, experiment results on an industrial CAPL are obtained to demonstrate the effectiveness of the proposed method.

The continuous annealing process line (CAPL) of cold rolling is an important unit to improve the mechanical properties of steel strips in steel making. In continuous annealing processes, strip tension is an important factor, which indicates whether the line operates steadily. Abnormal tension profile distribution along the production line can lead to strip break and roll slippage. Therefore, it is essential to estimate the whole tension profile in order to prevent the occurrence of faults. However, in real annealing processes, only a limited number of strip tension sensors are installed along the machine direction. Since the effects of strip temperature, gas flow, bearing friction, strip inertia, and roll eccentricity can lead to nonlinear tension dynamics, it is difficult to apply the first-principles induced model to estimate the tension profile distribution. In this paper, a novel data-based hybrid tension estimation and fault diagnosis method is proposed to estimate the unmeasured tension between two neighboring rolls. The main model is established by an observer-based method using a limited number of measured tensions, speeds, and currents of each roll, where the tension error compensation model is designed by applying neural networks principal component regression. The corresponding tension fault diagnosis method is designed using the estimated tensions. Finally, the proposed tension estimation and fault diagnosis method was applied to a real CAPL in a steel-making company, demonstrating the effectiveness of the proposed method.