Hierarchical Self-Attention Network for Industrial Data Series Modeling With Different Sampling Rates Between the Input and Output Sequences.

For industrial processes, it is significant to carry out the dynamic modeling of data series for quality prediction. However, there are often different sampling rates between the input and output sequences. For the most traditional data series models, they have to carefully select the labeled sample sequence to build the dynamic prediction model, while the massive unlabeled input sequences between labeled samples are directly discarded. Moreover, the interactions of the variables and samples are usually not fully considered for quality prediction at each labeled step. To handle these problems, a hierarchical self-attention network (HSAN) is designed for adaptive dynamic modeling. In HSAN, a dynamic data augmentation is first designed for each labeled step to include the unlabeled input sequences. Then, a self-attention layer of variable level is proposed to learn the variable interactions and short-interval temporal dependencies. After that, a self-attention layer of sample level is further developed to model the long-interval temporal dependencies. Finally, a long short-term memory network (LSTM) network is constructed to model the new sequence that contains abundant interactions for quality prediction. The experiment on an industrial hydrocracking process shows the effectiveness of HSAN.

Quality Prediction Modeling for Industrial Processes Using Multiscale Attention-Based Convolutional Neural Network.

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.

Efficiency Maximization of a Direct Internal Reforming Solid Oxide Fuel Cell in a Two-Layer Self-Optimizing Control Structure.

Control system configuration is essential for the efficiency performance of a solid oxide fuel cell (SOFC). In this paper, we aim to report a novel two-layer self-optimizing control (SOC) system for the efficiency maximization of a direct internal reforming SOFC, where the efficiency index is defined as the profit of generated electricity penalized by carbon (CO2) emission. Based on the lumped-parameter model of the SOFC, comprehensive evaluations are carried out to identify the optimal controlled variables (CVs), the control of which at constant set-points can optimize the efficiency, in spite of operating condition changes. In the lower SOC layer, we configure single variables as the CVs. The results show that the stack temperature is the active constraint which should be controlled to maintain the cell performance. In addition, the outlet hydrogen composition is identified as the optimal CV. This result differs from several previous proposals, such as methane composition. In the presence of operating condition changes, the set-point of hydrogen composition is further automatically adjusted by the upper SOC layer, where a linear combination of the SOFC measurements is configured as the CV, giving negligible efficiency losses. The cascaded two-layer SOC structure is able to maximize the SOFC efficiency and reduce carbon emission without using online optimization techniques; meanwhile, it allows for smooth and safe operations. The validity of the new scheme is verified through both static and dynamic evaluations.

Enhanced control structure design for an industrial off-gas system: Simple reconfigurations benefit the economy.

For control structure design of the industrial off-gas benchmark system, application of the Skogestad's state-of-art design procedure has suggested the scrubber inlet pressure (P si in the roaster) and one of the fan speeds (N fan1 or N_fan2 in the furnace) as the self-optimizing controlled variables CVs. In this study, we stress and advocate the gSOC-plus-BAB approach as an enhanced design toolkit for the classical and systematical design procedure. The gSOC (global self-optimizing control) is able to efficiently solve measurement combinations as CVs with improved economic performances, while the BAB (branch and bound) algorithm serves to fast screen promising measurement subsets for large-scale problems. Using the enhanced design for the off-gas system, our new findings are to control the combination of roaster's ID fan outlet pressures, 0.494P IDfan1+0.506P IDfan2 (setpoint: -257.04 Pa), and the furnace's fan pressure difference, P IDfan1- P IDfan2 (setpoint: 0). Such simple reconfigurations can dramatically reduce the average economic loss by 53.3% for the roaster and even achieve perfect optimal control for the furnace. Both steady state and dynamic evaluations are carried out to validate the reconfigured control structures. © 2017 Elsevier Inc. All rights reserved.

Real-Time Path Planning for Robot Using OP-PRM in Complex Dynamic Environment.

During task execution, the autonomous robots would likely pass through many narrow corridors along with mobile obstacles in dynamically complex environments. In this case, the off-line path planning algorithm is rather difficult to be directly implemented to acquire the available path in real-time. Hence, this article proposes a probabilistic roadmap algorithm based on the obstacle potential field sampling strategy to tackle the online path planning, called Obstacle Potential field-Probabilistic Roadmap Method (OP-PRM). The obstacle potential field is introduced to determine the obstacle area so as to construct the potential linked roadmap. Then the specific range around the obstacle boundary is justified as the target sampling area. Based on this obstacle localization, the effectiveness of the sampling points falling into the narrow corridors can be increased greatly for feasible roadmap construction. Furthermore, an incremental heuristic D* Lite algorithm is applied to search the shortest paths between the starting point and the target point on the roadmap. Simulation experiments demonstrate that the OP-PRM path planning algorithm can enable robots to search the optimal path fast from the starting point to the destination and effectively cross narrow corridors in complex dynamic environments.

Nonlinear Dynamic Soft Sensor Development with a Supervised Hybrid CNN-LSTM Network for Industrial Processes.

A soft sensor is a key component when a real-time measurement is unavailable for industrial processes. Recently, soft sensor models based on deep-learning techniques have been successfully applied to complex industrial processes with nonlinear and dynamic characteristics. However, the conventional deep-learning-based methods cannot guarantee that the quality-relevant features are included in the hidden states when the modeling samples are limited. To address this issue, a supervised hybrid network based on a dynamic convolutional neural network (CNN) and a long short-term memory (LSTM) network is designed by constructing multilayer dynamic CNN-LSTM with improved structures. In each time instant, data augmentation is implemented by dynamic expansion of the original samples. Moreover, multiple supervised hidden units are trained by adding quality variables as part of the layer input to acquire a better quality-related feature learning performance. The effectiveness of the proposed soft senor development is validated through two industrial applications, including a penicillin fermentation process and a debutanizer column.

Quality-Driven Regularization for Deep Learning Networks and Its Application to Industrial Soft Sensors.

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.

To Procrastinate or Not to Procrastinate: A Retrospective Study of the Optimal Timing of Containing the Global Spread of the COVID-19 Pandemic.

As global public health is under threat by the 2019-nCoV and a potential new wave of large-scale epidemic outbreak and spread is looming, an imminent question to ask is what the optimal strategy of epidemic prevention and control (P&C) measures would be, especially in terms of the timing of enforcing aggressive policy response so as to maximize health efficacy and to contain pandemic spread. Based on the current global pandemic statistic data, here we developed a logistic probability function configured SEIR model to analyse the COVID-19 outbreak and estimate its transmission pattern under different "anticipate- or delay-to-activate" policy response scenarios in containing the pandemic. We found that the potential positive effects of stringent pandemic P&C measures would be almost canceled out in case of significantly delayed action, whereas a partially procrastinatory wait-and-see control policy may still be able to contribute to containing the degree of epidemic spread although its effectiveness may be significantly compromised compared to a scenario of early intervention coupled with stringent P&C measures. A laissez-faire policy adopted by the government and health authority to tackling the uncertainly of COVID19-type pandemic development during the early stage of the outbreak turns out to be a high risk strategy from optimal control perspective, as significant damages would be produced as a consequence.

Retrospect: The Outbreak Evaluation of COVID-19 in Wuhan District of China.

There were 27 novel coronavirus pneumonia cases found in Wuhan, China in December 2019, named as 2019-nCoV temporarily and COVID-19 formally by the World Health Organization (WHO) on the 11 February 2020. In December 2019 and January 2020, COVID-19 has spread on a large scale among the population, which brought terrible disaster to the life and property of the Chinese people. In this paper, we analyze the features and pattern of the virus transmission. Considering the influence of indirect transmission, a conscious-based Susceptible-Exposed-Infective-Recovered (SEIR) (C-SEIR) model is proposed, and the difference equation is used to establish the model. We simulated the C-SEIR model and key important parameters. The results show that (1) increasing people's awareness of the virus can effectively reduce the spread of the virus; (2) as the capability and possibility of indirect infection increases, the proportion of people being infected will also increase; (3) the increased cure rate can effectively reduce the number of infected people. Then, the virus transmission can be modelled and used for the inflexion and extinction period of pandemic development so as to provide theoretical support for the Chinese government in the decision-making of pandemic prevention and recovery of economic production. Further, this study has demonstrated the effectiveness of the prevention measures taken by the Chinese government such as multi-level administrative district isolation and public health awareness.

The global COVID-19 pandemic at a crossroads: relevant countermeasures and ways ahead.

BACKGROUND: Since the outbreak of novel coronavirus disease (COVID-19) in Wuhan, China at the beginning of December 2019, there have been over 11,200,000 confirmed cases in the world as of the 3rd July 2020, affecting over 213 countries and regions with nearly 530,000 deaths. The pandemic has been sweeping all continents, North America, Latin America, Europe, Middle East and South Asia among others at an alarming rapidity. Here, we provide an estimate of the scale of the pandemic spread under different scenarios of variation in key influencing parameters with a hybrid model. METHODS: We developed a new hybrid model of infectious disease transmission based on Cellular Automata (CA)-configured SEIR to analyse the COVID-19 outbreak and estimate its transmission pattern. A probabilistic contamination network is embedded in the pandemic transmission model to capture the randomness feature of person-to-person spread of the novel virus. We used the improved SEIR model to quantify the population contact state with isolation measures under different continuous time series contact probability via CA. We adjusted the modelling parameters to verify the model performance in accordance to the data from the reports published by the Chinese Center for Disease Control and Prevention. We simulated several scenarios by varying such key parameters as number of isolation rate, average contact times of the population, number of infected people before taking prevention and control measures, medical level and number of imported cases. RESULTS: In the baseline model, we identified that the isolation control as the most influencing factor that had the largest impact on decreasing the speed of the reproductive number, accelerating the arrival of the "inflection point" of pandemic prevention and control, and the death rate reduction. We estimated that the probability of people contacts and the number of the onset infected cases before prevention measures also had significant effect on the infection rate reduction with appropriate prevention measures adoption, which partly reflects the impact of timely measure on the severity of the outbreak. We found that imported cases will risk the domestic prevention. CONCLUSIONS: Our modelling results clearly indicate that early-stage preventive measures are the most effective way to contain the pandemic spread and a strong interventionist approach needs to be adopted by policymakers vis-à-vis of the highly contagious nature of the COVID-19. Human resources, intensified isolation and confinement as well as special hospital buildings should be prioritised in countries with large number of infections to constrain the global transmission of the virulent infection. To do so, internationally coordinated actions require to be taken to replicate good practices to less infected countries and regions immediately.