Search

Stable fault self-healing ensures the smooth operation of intelligent industrial manufacturing processes. However, current research, while capable of inferring fault propagation paths or individually controlling manufacturing processes, still faces some challenges and limitations in addressing issues within manufacturing processes. In particular, it is easy to overlook the impact of fault variables on normal causal relationships in the manufacturing process and how to achieve adaptive fault repair from the source of faults. The biological immune system possesses sensitive nodes and a neuro-endocrine-immune network with interconnected functionalities, bearing remarkable resemblance to the sensor arrays, information networks, and intelligent controller systems in modern self-healing control systems. These problem domains have significant practical implications in industrial manufacturing processes and urgently require new methods and technologies to address them. To address the issue of root cause fault self-healing in intelligent industrial manufacturing processes, this paper proposes an innovative method that utilizes root cause inference and bio-inspired heuristic reinforcement learning models to achieve fault self-healing in industrial processes. First, innovatively introduce the Variational Autoencoder (VAE) into the Time Convolutional Network (TCN) to construct the TCN-VAE network. By performing feature reconstruction, the network's feature extraction capability is enhanced, further exploring the relationships between latent variables, and consequently building a causal graph of faults. A multi-head attention mechanism is introduced into the network, and the inference process is quantitatively evaluated, thereby improving the generalization and accuracy of root cause inference. Root cause determination rules are established to identify the fundamental reasons for fault occurrence, ensuring self-healing from the source of faults. Next, we developed for the first time a fault self-healing model based on reinforcement learning, utilizing the immune repair process of effector T cells to characterize the fault self-healing process. It effectively adapts to eliminate various levels of faults based on the state transition process of antigens in T helper cells. Additionally, stability control analysis is performed on the proposed model. The average reward function curve and Sobol sensitivity index demonstrates the model's strong robustness in practical applications. The validation results from case studies on the Tennessee Eastman (TE) and Continuous Stirred Tank Reactor (CSTR) show that the average improvement in fault repair F1 score was 0.0941 and 0.105 respectively, and the fastest improvement in fault repair response time was 0.266s and 0.258s respectively. It was confirmed that the root cause inference and fault repair results align with the actual mechanisms of the manufacturing processes. The proposed method has been verified to achieve stable and accurate root cause fault repair in real-world intelligent industrial manufacturing systems, making a significant contribution to achieving stable and accurate root cause fault repair in intelligent industrial manufacturing systems. • TCN-VAE enables feature reconstruction and latent variable exploration for root cause reasoning. • Multi-head attention enhances generality and accuracy in root cause reasoning. • Effector T-cell immune-inspired RL adapts to eliminate faults effectively. • The fault self-healing model exhibits strong reliability and robustness. • Method validated on Tennessee Eastman and continuous stirred-tank reactor datasets. [ABSTRACT FROM AUTHOR]

Adoptive T cell therapy has undergone remarkable advancements in recent decades; nevertheless, the rapid and effective ex vivo expansion of tumor-reactive T cells remains a formidable challenge, limiting their clinical application. Artificial antigen-presenting substrates represent a promising avenue for enhancing the efficiency of adoptive immunotherapy and fostering T cell expansion. These substrates offer significant potential by providing flexibility and modularity in the design of tailored stimulatory environments. Polydimethylsiloxane (PDMS) silicone elastomer stands as a widely utilized biomaterial for exploring the varying sensitivity of T cell activation to substrate properties. This paper explores the optimization of PDMS surface modification and formulation to create customized stimulatory surfaces with the goal of enhancing T cell expansion. By employing soft PDMS elastomer functionalized through silanization and activating agent, coupled with site-directed protein immobilization techniques, a novel T cell stimulatory platform is introduced, facilitating T cell activation and proliferation. Notably, our findings underscore that softer modified elastomers (Young' modulus E∼300 kPa) exhibit superior efficacy in stimulating and activating mouse CD4+ T cells compared to their stiffer counterparts (E∼3 MPa). Furthermore, softened modified PDMS substrates demonstrate enhanced capabilities in T cell expansion and Th1 differentiation, offering promising insights for the advancement of T cell-based immunotherapy. [Display omitted] • Chemically modified layers and Protein A enhance antibody density and stability. • Softer elastomers better stimulate CD4+ T cells compared to stiffer ones. • Softened PDMS substrates enhance T cell expansion and Th1 differentiation. • Study shows surface chemistry and mechanical cues influence T cell behavior. [ABSTRACT FROM AUTHOR]

Highly expressed azo-reductase (AZOR) selectively reduces Azo-DON to DON in hypoxic tumor environments, impair glutamine metabolism in cancer cells, and promote tumor cell death without affecting T cell metabolism and proliferation, thereby generating strong antitumor immune responses. CB-PLG nanoparticles (CBP) mentioned in the paper can enhance the degree of hypoxia in normoxic tumors, which is conducive to efficient reduction of Azo-DON at tumor sites. [Display omitted] • Azo-DON is a novel hypoxia-active prodrug of glutamine antagonist. • Azo-DON selectively blocks tumor glutamine metabolism after reducing to DON. • T cells continue to proliferate and remain active when Azo-DON is reduced to DON. • CBP nanoparticles increase tumor hypoxia and facilitate the reduction of Azo-DON. Glutamine antagonists, such as 6-diazo-5-oxo-L-norleucine (DON), have demonstrated remarkable anti-tumor effects by blocking tumor glutamine metabolism, but their use is frequently causing toxicity. Despite the existence of multiple glutamine antagonist prodrug designs, a tumor-selective prodrug has yet to be developed. Herein, a novel prodrug of DON, Azo-DON, has been developed, which remains stable and inactive in normal tissues with sufficient oxygen levels, while can be selectively reduced to DON by highly expressed azo-reductase in hypoxic tumor environments. This leads to blockade of glutamine metabolism in cancer cells and promotes cell death without affecting T cell proliferation. In a high-hypoxic H22 hepatoma cancer model, Azo-DON showed a 1.8-fold enhancement in glutamine blockade compared to the control group, resulting in a tumor suppression rate (TSR) of 84.2% in vivo with no significant weight loss. In a low-hypoxic CT26 colon cancer model, when combined with vascular disrupting agent nanoparticles (CBP) to induce a hypoxic environment, Azo-DON exhibited a 4.6-fold enhancement in glutamine blockade over the control group, resulted in a remarkable TSR of 96.6% in vivo. This innovative approach represents a promising strategy for the application of broad-spectrum metabolic inhibitors in the field of precision cancer treatment. [ABSTRACT FROM AUTHOR]