With the rapid development of artificial intelligence (AI) in recent years, fault diagnostics for industrial applications have leaped toward partially or fully automatic provided by the capability of analyzing massive condition monitoring data from sensors and actuators. Generally, AI-based fault diagnostics can achieve high accuracy when failure types appear in training dataset and testing dataset are the same. These diagnostic methods could be invalidated for applications dealing with unprecedented faults because the pretrained classifier for diagnostics tends to misclassify the novel instances into existing known classes. In order to address these limitations of conventional diagnostic approaches, we propose a unified diagnostics framework that can achieve novel fault detection and known fault classification tasks together. Through jointly training a variational autoencoder and a deep neural networks classifier, we convert the original entangled raw data into latent variables with Gaussian probabilistic distributions in the latent space and utilize the probabilistic latent variables to detect novel samples against known fault classes or classify them into one of the existing fault classes if they are not novel. The effectiveness of our proposed joint-training framework is validated through experimental studies on two different bearing datasets. Compared with the state-of-the-art methods in the literature, our unified framework is able to not only accurately detect the novel fault classes but also achieve high classification accuracy of known fault classes.

In recent years, driven by Industry 4.0 wave, academic research has focused on the science, engineering, and enabling technologies for intelligent and cyber manufacturing. Using a network science and data mining-based Keyword Co-occurrence Network (KCN) methodology, this work analyzes the trends in data science topics in the manufacturing literature over the past two decades to inform the researchers, educators, industry leaders of knowledge trends in intelligent manufacturing. It studies the evolution of research topics and methods in data science, Internet of Things (IoT), cloud computing, and cyber manufacturing. The KCN methodology is applied to systematically analyze the keywords collected from 84,041 papers published in top-tier manufacturing journals between 2000 and 2020. It is not practically feasible to review this large body of literature through tradition manual approaches like systematic review and scoping review to discover insights. The results of network modeling and data analysis reveal important knowledge components and structure of the intelligent and cyber manufacturing literature, implicit the research interests switch and provide the insights for industry development. This paper maps the high frequency keywords in the recent literature to nine pillars of Industry 4.0 to help manufacturing community identify research and education directions for emerging technologies in intelligent manufacturing.

Traffic signal control (TSC) is a challenging problem within intelligent transportation systems and has been studied using multi-agent reinforcement learning (MARL). While centralized approaches are often infeasible for large scale TSC problems, decentralized approaches provide scalability but introduce new challenges, such as partial observability. Communication plays a critical role in decentralized MARL as agents must learn to encode information in messages to better understand the system and achieve effective coordination. Deep MARL has been used to enable inter-agent communication by learning communication protocols in a differentiable manner. However, many deep MARL communication frameworks proposed for TSC allow agents to communicate with all other agents at all times, which can add to the existing noise in the system and degrade the overall performance. In this study, we propose a communication-based MARL framework for large-scale TSC. Our framework allows each agent to learn a communication policy that dictates "which" part of the message is sent "to whom". This results in a decentralized and flexible communication mechanism where agents can effectively use the communication channel only when necessary. We designed two networks, a synthetic 4 × 4 grid network and a real-world network based on the Pasubio neighborhood in Bologna. Our framework achieved the lowest network congestion compared to related methods, with agents utilizing ~ 47 – 65% of the communication channel. Ablation studies further demonstrated the effectiveness of the communication policies learned within our framework.

Roll-to-roll (R2R) manufacturing has emerged as one of the most cost-effective manufacturing techniques for high-volume production of flexible printing electronics, graphene films, thin-film batteries, and solar cells, as opposed to low-volume batch production. However, in practice, real-time monitoring of the R2R process and in situ measurement of fabricated materials during production are challenging because of the high speed and dynamic variation of the web. In addition, due to the lack of cost-effective sensors and in-line metrology systems with a large range and high resolution, effective quality control and defect diagnosis are difficult to achieve. To address these challenges, this paper aims to develop analytical models that can serve as virtual sensing and metrology tools to quantify process state variation and estimate product quality in R2R processes with limited accessibility of in situ sensor. First, the quality variation propagation mechanism in R2R processes is investigated. Second, a hybrid multistage modeling method is proposed to characterize the twofold variation propagation--product-centric and process-centric variations, and its relationship with product quality in R2R processes.