Stable operation of the sintering process is critical to ensuring the final quality of ternary cathode materials. However, the strong nonlinearity and dynamics, resulting from the process feedback control, coupled temperature zones, and the unstable external environment, often lead to the traditional process monitoring methods monitoring poor performance. To this end, this paper proposes a recurrent variational auto-encoder (RVAE)-based industrial process monitoring method. First, the variational auto-encoder-based nonlinear dynamic system (NDS) model of the sintering process is learned, which can fully extract the process nonlinearity. Then, the autoregressive equation among the latent variables is established according to a recurrent neural network, and the weights to samples at various times are assigned by the weighted moving average method. On this basis, the dynamic characteristics among the latent variables can be excavated effectively. Subsequently, some appropriate monitoring statistics are designed based on principal component analysis (PCA), which are more sensitive to faults than indicators based on reconstruction errors. Correspondingly, the RVAE-based process monitoring strategy is proposed. Finally, the proposed method has been verified to be effective and superior by the industrial application of the sintering process.
2 September 2025 in IEEE Transactions on Cognitive and Developmental Systems
Mobile manipulators need to determine feasible navigation positions before manipulation tasks. Real-world environments, with varying obstacles and objects, pose significant challenges for computing optimal navigation positions due to their variability. In this work, a novel method named Graph Reinforcement Learning-based Reachability Map (GRAM) is proposed. First, GRAM uses a graph attention network (GAT) to capture the spatial relationships between objects. Then, it leverages the Q-value from the pre-trained critic network to generate the reachability map. The reachability map is integrated into navigation policies for long-horizon tasks, effectively solving the skill transition problems. Extensive simulation and real-world experiments were conducted on the Fetch mobile robot platform. The results demonstrate the superiority of GRAM, with simulation results showing an average 16.3% performance improvement over the baseline in four flexible environments. In long-horizon tasks, GRAM’s overall task success rate improved by 4.2%. The project is open-source at https://github.com/nubotnudt/Grand RM.